The Industrial Pollutionm

THE INDUSTRIAL-POLLUTION-PROJECTION SYSTEM:
CRITICAL ANALYSIS AND POTENTIAL APPLICATIONS

by
PAUL J. MARTIN
B.A.(Hons) Geography and Management Studies
Cambridge University, England
(1987)
SUBMITTED TO THE DEPARTMENT OF URBAN

STUDIES AND PLANNING IN PARTIAL EULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER IN CITY PLANNING
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
May 1993
@ Paul J. Martin 1993

All rights reserved
The author hereby grants to M.I.T.
permission to reproduce and to distribute publicly copies
of this thesis document in whole or in part
Signature of Author
Departint ofl rban Studies and Planning
Certified by
Karen R. Polenske
Profess r of Regional Political Economy and Planning
Thesis Supervisor
Accepted by_
Ralph A. Gakenheimer
Professor of Urban Planning
Head, Master of City Planning Committee
Rotch
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
[JUN 03 1993
UBRARIES
- 2 THE INDUSTRIAL-POLLUTION-PROJECTION SYSTEM:

CRITICAL ANALYSIS AND POTENTIAL APPLICATIONS
by
Paul J. Martin
Submitted to the Department of Urban Studies and Planning
on May 20, 1993 in partial fulfillment of the requirements
for the Degree of Master in City Planning
ABSTRACT
Rapid industrialization is forcing policymakers in many developing
countries to face the challenge of industrial-pollution control. The
role of government in correcting this market failure is to improve net
social welfare by reducing pollution until the marginal cost of
pollution control equals the marginal benefit. To achieve any success
in this complex task, policymakers require a substantial amount of
information and strong institutional capacity. Unfortunately, both
these requirements are in short supply in most developing countries.
The need for a planning tool to assist in the formulation of
industrial-pollution-control policy was recognized by the World Bank in
1990 and led to the development of the Industrial-Pollution-Projection
System (IPPS) using U.S. Environmental Protection Agency and U.S. Census
of Manufactures data. By linking estimates of pollution intensity to
sectoral production data, policymakers can use IPPS to assist in the
identification of key pollutants and priority sectors, and to provide an
important input to their assessments of the costs and benefits of
pollution-control policy.
IPPS is based on a set of sectoral pollution-intensity
coefficients drawn from a sample of U.S. manufacturing plants. As such,
some error is to be expected in IPPS estimates of pollution loads, even
using U.S. data. In this thesis I present some preliminary evidence
that these errors may be significant. A substantial amount of work is
required to identify the sources of these errors and to determine if any
improvements can be made. I conclude that the potential value of the
system to planners in the World Bank and client governments is
sufficient to justify the additional effort.
Thesis Supervisor: Professor Karen R. Polenske

Title: Professor of Regional Political Economy and Planning
- 3 ACKNOWLEDGEMENTS
I wish to express my gratitude and appreciation to all those who

have helped in this research.
Many people gave freely of their time and
energy to provide the necessary data, for which I would particularly

like to thank Dr. Barry Wallerstein and Francis Goh of the South Coast
Air Quality Management District, Professor George Treyz of Regional
Economic Models Inc., Bob McGuckin and Arnie Reznek of the Center for
Economic Studies at the Census Bureau, and Manjula Singh at the World
Bank.
I would like to record my special thanks to Professor Karen
Polenske, a devoted educator, for her patient supervision of this work
and her many helpful suggestions.
Credit for the creation of the
project must go to David Wheeler at the World Bank, who has been a
constant source of inspiration throughout, and who has provided much
important feedback.
Valuable criticism and counsel have also come from
Professor Lawrence Susskind, for which I am very grateful.

Finally, I would like to thank my parents for providing me with
the opportunity to conduct this research, and Maria Vazquez Torres for
providing perspective.
- 4 TABLE OF CONTENTS
Abstract. .................................................................. 2
Acknowledgements........................................................ 3
Table of Contents......................................................
Chapter 1.
INTRODUCTION: THE INDUSTRIAL-POLLUTION-POLICY IMPERATIVE ..............
Obstacles to Effective Industrial-Pollution-Control Policy...... 7

Thesis Structure................................................ 10
Chapter 2.
ADDRESSING ENVIRONMENTAL POLICY ISSUES................................. 11
Identifying Key Pollutants and Priority Sectors: An Illustration
Using Indonesian Data....................................
14
Assessing the Costs of Industrial-Pollution-Control Policies... 22

Determining the Benefits of Industrial-Pollution Control....... 31
Importance of Developing Tools for Industrial-Pollution
Policy-Making............................................ 34
Chapter 3.
INITIAL DEVELOPMENT. ..................................................... 36
Conceptual Goal................................................ 38
Building Blocks for IPPS........................................ 40
EPA Emissions Databases.................................. 40
Longitudinal Research Database............................. 44
Construction of Pollution-Intensity Indices..................... 45
- 5 Choice of Numerator...................................... 46
Choice of Denominator.................................... 50
Results........................................................ 52
Chapter 4.
CRITICAL ASSESSMENT AND FURTHER WORK................................... 54
IPPS Indices...................................................
55
Sources of Bias.................................................. 61
Statistical Confidence in the IPPS Indices...................... 62
Testing the IPPS Estimates at the National and the
Regional Level............................................. 67
IPPS Estimates of National Emissions...................... 67
IPPS Estimates for the South Coast Air Quality
Management District.................................. 71
Further Work................................................... 78
Bibliography..........................................................
82
Glossary of Abbreviations............................................ 84
- 6 CHAPTER 1
INTRODUCTION: THE INDUSTRIAL-POLLUTION-POLICY IMPERATIVE
In Bangkok, Beijing, Calcutta, Jakarta, Tehran, and other urban

centers in developing countries around the world, some 1.3 billion
people breathe air that does not meet World Health Organization
standards (World Bank, 1992).
The resulting sickness and death is worse
than would be expected under the same conditions in developed nations,

because of pre-existing poor health and nutrition.
Between 300,000 and
700,000 premature deaths are caused each year from particulate pollution
alone (World Bank, 1992).
Other priority pollution issues in the
world's rapidly industrializing regions include the exposure of more

than a billion people to unhealthy levels of sulfur dioxide and lead,
loss of welfare from water-systems that have become biologically dead
from the release of organic pollutants, and sickness caused by the
uncontrolled disposal of toxic wastes.
The problems of urban pollution present the World Bank with one of
its greatest challenges in its mission to "alleviate poverty and promote
sustainable development" (World Bank, 1992).
In this thesis, I examine
the development by the World Bank of the Industrial-Pollution-Projection

System (IPPS), a tool designed to assist in the formulation and
implementation of industrial-pollution-control policy.
I assess the
importance of such a tool for improving the quality of environmental

decision-making, then I describe the development of IPPS, and finally I
critically examine the validity of the system.
In order to place these
issues in some context, I outline below some of the obstacles facing

governments and development agencies in establishing and enforcing
- 7 effective industrial-pollution policy in developing countries.

Obstacles to Effective Industrial-Pollution-Control Policy
Industrial-pollution-control policy may be characterized as
government intervention to correct a market failure.
generated as an externality to industrial production.
Pollution is
The market
failure arises because incomplete information and transaction costs

prevent the full social cost of pollution from being reflected in the
production decision of the generating agent.
welfare is not optimized.
As a result, net social
The social costs of pollution take a number
of forms, which may be summarized in three broad categories as (1)

detriments to human health, (2) loss of productive capacity, and (3)
loss of aesthetic utility.
Although detriments to human health may also
cause a reduction in productive capacity, they are defined separately

because some health effects may reduce welfare without directly
affecting production.
The definition of lost productive capacity
includes damage to both physical capital and natural resources.
Lost
aesthetic utility includes the many aspects of individual enjoyment of

the natural and built environment that may be adversely affected by
industrial pollution, but which are difficult to quantify in monetary
Assessment of all these costs is further complicated by the fact
terms.
that current pollution may reduce social welfare far into the future,
raising the problems of social and intergenerational discounting.
Simply stated, the objective of pollution-control policy is to
improve net social welfare by reducing pollution until the marginal
social cost of pollution control is equal to the marginal social
benefit.
In practice, the assessment of both the costs and benefits is
- 8 difficult.
The controversy surrounding this task is highlighted by an
infamous memorandum leaked from the Chief Economist of the World Bank,
Lawrence Summers.
Despite the well-recognized detrimental impacts of
industrial pollution outlined above, Summers wrote:
Just between you and me, shouldn't the World Bank be encouraging
more migration of the dirty industries to the LDCs ? (Summers,
1992)
Understandably, this memorandum caused an international uproar.

Nevertheless, behind the blunt language lurks the serious issue of the
appropriate level of pollution control in lower-income nations.
Difficulties arise for policy-makers in developing countries
largely because there are good grounds to believe that neither the
benefits of industrial production nor the costs of industrial pollution
match those in the developed world.
Perhaps the most obvious social
costs of pollution are those associated with adverse impacts on human

health.
Just as the marginal detriment to health is likely to be higher
in developing countries because of pre-existing ill-health and

malnutrition, so the marginal improvement in health from an increase in
national income is likely to be greater than in the developed world.
Possibly the most objectionable aspect of Summers' memorandum was the
suggestion that the costs of health-impairing pollution depend on the
foregone earnings from increased morbidity and mortality.
As Jagdish
Bhagwati wrote in response:

. . . no modern economist, when his house is on fire, will pull
his father out before his mother because the father earns more
than the mother. Economists have learned, at least since the
1960's, to broaden their analysis to include objectives (including
the environment) other than just goods and services. (Bhagwati,
1992)
- 9 If part of the aim of industrial-pollution control is to improve

human health, then the human health costs of reduced national income,
caused by the increased production costs associated with pollution
control, should also be taken into account.
By converting the abatement
costs into a measure of the impact on human health, they can be directly
compared with the human health benefits of pollution control.
This form
of analysis is complex, and it can be conducted only in approximate

terms, but it is essential to do if pollution-control policies in lowerincome nations are to achieve efficient outcomes.
The two essential ingredients for the implementation of effective
policies to control industrial pollution are information and
institutional capacity.
Without reliable information, policy-makers
cannot begin to determine which policies will deliver the most efficient
outcomes.
Without sufficient institutional capacity, the information
cannot be effectively analyzed and the policies cannot be implemented.

Both of these constraints are more severe in developing countries than
in the industrialized world, and both can crucially benefit from the
availability of effective planning tools.
The development of IPPS by
the World Bank was undertaken in recognition of these needs.
First,
IPPS is intended to allow better-informed policy-making by providing

estimates of pollution emissions from widely available economic data,
where emissions data was previously scarce.
Second, IPPS is intended to
strengthen institutional capacity in two ways.
The system will allow
more efficient use to be made of scarce financial and administrative

resources by helping to identify priority pollutants and sources.
addition, by providing a clear framework for setting industrial-
In
10
pollution-control policy, IPPS will improve the transparency of

decision-making.
This is especially important for government and
development agencies facing political pressure in this arena.
Thesis Structure
In the next chapter of this thesis, I discuss in more detail the
potential value of a tool such as IPPS for improving the quality of
environmental decision-making, especially for developing countries.
First, I illustrate how the system may be used to identify key
pollutants and priority sectors.
I then demonstrate how IPPS may be
used in conjunction with economic models to assess the costs of an

environmental policy, and, finally, I discuss the potential contribution
of the tool in modeling ambient conditions.
In Chapter 3, I describe
the conceptual goal of the World Bank team who produced the initial
version of IPPS, the methodology we employed, and some of the
operational complexities we encountered.
I critically assess the
validity of our results in the Chapter 4, indicating the existence of

some serious problems.
I end by suggesting some further work that needs
to be undertaken to improve the accuracy and utility of IPPS, arguing

that the additional effort is justified by the potential contribution of
such a tool to the formulation of industrial-pollution-control policies.
11
CHAPTER 2
IMPROVING ENVIRONMENTAL POLICY-MAKING
In the introductory chapter, I outlined the obstacles to the

formulation and implementation of effective environmental policy in
developing countries.
Here, I demonstrate more explicitly the ways in
which a tool such as the Industrial-Pollution-Projection System (IPPS)

may help overcome these obstacles.
In simplest terms, IPPS is a set of
pollution-intensity coefficients.
Each coefficient describes the amount
and type of pollution that may be expected as an externality to

production in a given industrial sector.
Thus, IPPS is able to create
estimates of the pollution load generated by a given set of

manufacturing installations.
In order to define more precisely the
potential contributions and limitations of the system, I frame this

discussion in the terms used by the World Bank, in the World Development
Report 1992, "Development and the Environment" (World Bank, 1992).
Chapter Four of the report is particularly salient, providing in the
chapter-heading a basis for examining the potential of IPPS; "Making
Better Decisions: Information, Institutions, and Participation."
Although my main argument in this chapter is that IPPS has the
potential to improve environmental decision-making significantly through
providing better information, I want to note at the outset that such a
tool may also contribute towards institutional-strengthening and
consensus-building.
According to the World Development Report 1992,
institutional-strengthening includes the need to develop legislation and

administrative structures, provide skills, and ensure adequate funding.
12
Clearly, IPPS offers the potential to provide skills to the extent that
such a tool enhances the technical capacity of an institution.
Beyond
this, better identification of priority pollutants and sources may also

aid in the establishment of a more effective administrative structure,
and a more efficient allocation of funds than is possible in the absence
of such information.
I present this argument in more detail in the
first section below.
Turning to the issue of participation, although
the availability of an effective planning tool does not directly promote

public involvement, it provides agencies with another mechanism for
achieving consensus.
A clearly established set of analytical techniques
helps agencies to shift the debate away from political interests, to a

more technical level of dialogue, at which acceptance of methodology and
assumptions substantially reinforces acceptance of eventual policy
conclusions.
Nevertheless, IPPS is primarily designed to raise the quality of
information available for environmental decision-making in developing
countries.
In discussing the importance of information for the
establishment of sound environmental policy, the World Development

Report 1992 states:
Ignorance is an important cause of environmental damage and a

serious impediment to finding solutions. .
. It is necessary,
first, to know the facts; second, to determine values and analyze

the benefits and costs of alternative measures; and, third, to
ensure that information is available to inform public and private
choices. (World Bank, 1992, p. 85).
IPPS has a potentially important role in meeting the first two of these
three necessities: the establishment of the facts and the analysis of
costs and benefits.
In many developing countries directly monitored
13
industrial-pollution data are scarce and expensive to obtain, so that

the facts of industrial pollution are poorly understood.
In the first
section below, I demonstrate the use of IPPS to estimate industrial

pollution from economic data.
This use of the system assists analysts
in the preliminary task of establishing the total pollution load from

manufacturing and, subsequently, in identifying the major source
sectors.
To aid in the illustration of the types of policy issue that
may be analyzed, I use economic data from Indonesia, one of the most
rapidly industrializing countries in the world.
The purpose of this
discussion is to underline the potential value of such a system;

however, analysts must treat the calculated pollution loadings with a
high degree of caution.
Although estimates of pollution loads are useful, they do not
directly identify the appropriate level of pollution control.
As I
mentioned in the previous chapter, the objective of pollution-control

policy is to improve net social welfare by forcing pollution abatement
until the marginal social cost of control equals the marginal social
benefit.
In the subsequent two sections, I discuss the role of IPPS in
helping to determine both the potential costs and the benefits of

proposed policies aimed at controlling industrial pollution.
This form
of analysis is particularly pertinent in developing countries, where we

do not know whether the direct transfer of western environmental
standards is likely to achieve optimal social outcomes.
As argued by
Tobey, the marginal social cost of raising the cost of industrial

production in a developing country is likely to be higher than in a
developed nation, because of the relatively greater impact on what he
14
terms "poverty-related" pollution (Tobey, 1989).
He uses this term to
describe the health impacts of poverty, for example, as a result of

limited access to sanitation and clean water, and a low standard of
nutrition.
To the extent that such "poverty-related" pollution has a
negative impact on human health, analysts can directly compare it to any

health improvements resulting from reduced industrial pollution.
In order to examine the costs associated with a proposed
industrial-pollution-control policy, analysts must link IPPS to a model
of the economy in question.
To examine more fully the policy issues
that may be raised, I again employ economic data, this time from
Massachusetts, in a simulation of a hypothetical environmental tax
introduced in the Commonwealth.
I chose this state because of the ready
availability of a highly sophisticated regional economic model, which

enables me to demonstrate the rich potential of this form of analysis.
In the third section of the chapter, I turn to the benefit-side of the
environmental policy calculus.
I discuss in more detail the relevance
and major components of a system in which analysts could use IPPS as one
input in assessing the health and welfare benefits of an industrialpollution-control policy.
Identifying Key Pollutants and Priority Sectors:

an Illustration using Indonesian Data
In this section, I demonstrate the relatively straight-forward use
of the IPPS intensities to establish total pollutant loadings and to
identify key source sectors.
I apply IPPS to Indonesian economic data
in order to demonstrate more clearly the environmental policy issues

that may be analysed.
As mentioned above, the use of Indonesian data is
15
for illustrative purposes, and little reliance may be placed on the

actual pollution loadings derived from the current version of IPPS.
The growth rate of manufacturing output in Indonesia reached 12.5%
during the 1980s, making it one of the most rapidly industrializing
countries in the world (World Bank, 1992).
Government policy-makers
have identified the need to control the associated growth in industrial

pollution as a priority issue in the preparation of Indonesia's second
25-year plan and the sixth 5-year plan (Repelita VI).
This has led
World Bank officials to establish a loan to strengthen the institutional

capacity for environmental management.
The lack of comprehensively
monitored data on industrial emissions or ambient conditions, however,

severely hampers analysts' efforts to identify priority pollutants and
emitters.
This makes the meaningful incorporation of environmental
issues into economic planning almost impossible.

IPPS offers the potential for analysts to use the relatively
comprehensive data from the Indonesian Census of Manufactures to
identify both the priority pollutants and the priority sectors for the
control of industrial pollution.
IPPS intensity coefficients are
available for about two-thirds of Indonesia's total industrial output in

the period 1976-1987 (see Figure 2.1).
For this example, I used the
census information and the IPPS intensities to estimate the total

release in Indonesia of four pollutants in 1976, 1980, 1984, and 1987.
The pollutants are carbon monoxide (CO), nitrogen dioxide (NO2 ),
particulates (PT), and sulphur dioxide (SO2 ).
I calculated the release
from each sector as the product of the recorded value of output and the
IPPS pollution intensity, and then I summed the sectoral data to derive
Figure 2.1
Indonesia - Industrial Production, 1976-1987
Total Production and Production from IPPS Sectors
(1987 $billion)
1976
1980
1984
Year
0 Totai ManufactuNng
+ IPPS Sectors
IPPS - Industrial Pollution Projection System

Source:
Industrial-Pollution-Projection System, World Bank
Indonesian Census of Manufactures, 1976-1987
1987
the total emissions.
17
By using this procedure, I ignored the pollution
from those sectors for which IPPS intensities are unavailable, amounting
to one-third of the total value of industrial output.
However, the
contribution of these sectors to the total pollution load is likely to

be small, because IPPS includes the most pollution-intensive industries.
Figure 2.2 charts the total estimated releases of the four
pollutants over the eleven-year period.
As can be seen, the annual
release of each pollutant increased approximately tenfold between 1976

and 1987.
The greatest increase was in the emission of CO, which jumped
from about 5,000 tons in 1976 to over 60,000 tons by 1987.
Although we
might expect most of this increase to be a direct result of the growth

in manufacturing output, it is also possible that the structure of the
industrial sector is becoming more pollution intensive.
Figure 2.3
illustrates this issue by charting the pollution intensity of

manufacturing output (calculated as the total release from manufacturing
activity divided by total manufacturing output) for the four pollutants
between 1976 and 1987.
This chart indicates that in 1987 every million
dollars of manufacturing output released more than three tons of CO,

compared to less than half that amount in 1976.
The intensity of
particulate emissions also almost doubled, while the intensity of NO 2

and SO2 emissions remained relatively stable.
The conclusion to be
drawn from this analysis is that the growth in output from the more CO
and PT intensive sectors exceeded that of the manufacturing sector as a
whole, whereas the NO 2 and SO2 emitting sectors grew at approximately
the same rate as total manufacturing output.
Figures 2.2 and 2.3 illustrate the potential of IPPS for
18
Figure 2.2
IPPS Estimates of Total Releases of Selected Pollutants
From Indonesian Manufacturing, 1976-1987
(Thousand Tons)
1976
1980
1984
1987
Year
CO
PT
N2
FS2
CO - Carbon Monoxide
IPPS -Industrial Pollution Projection System

N02 - Nitrogen Dioxide
PT - Total Particulates
S02 - Sulphur Dioxide
Source:
Industrial-Pollution-Projection System, Wold Bank

Figure 2.3
IPPS Pollution Intensity Estimates
Indonesian Manufacturing, 1976-1987
(Tons/$ million)
3.4.
1976
1980
1984
Year
OCc
+NO2
OPT
6So2
IPPS -Industrial Pollution Projection System
Source:

1987
identifying priority pollutants.
19
An equally important function of the
system is to identify the key polluting sectors.
In particular, in
order to establish and enforce effective environmental controls for a

rapidly industrializing economy, policy makers must gain some
understanding of the activities that contribute most to the increase in
pollution, such as the tenfold increase charted for Indonesia in Figure
2.2.
The sector-specific pollution intensities provided by IPPS allow
estimates to be made of the sectoral contributions to the total increase

in pollution.
Figures 2.4-2.7 present these estimates for the four
pollutants studied in this example.
The height of each bar represents
the percentage share of each sector in the total increase in emission of

each pollutant in Indonesia from 1976 to 1987.
The most important aspect of Figures 2.4-2.7 is the relatively
few sectors that account for almost the entire increase in the emission
of each pollutant.
Indeed, for each of the four pollutants illustrated,
the top nine sectors (the top quartile) account for at least 90% of the
increase.
In the case of CO emissions, the top quartile caused 98% of
the increase in emissions.

all four pollutants.
ISIC
ISIC
ISIC
ISIC
ISIC
ISIC
3118,
3211,
3311,
3411,
3512,
3710,
Six sectors appear in the top quartile for
These sectors are:
Sugar Factories and Refineries;

Spinning, Weaving and Finishing Textiles;
Sawmills, Planing and Other Wood Mills;
Pulp, Paper and Paperboard;
Fertilizers and Pesticides; and
Iron and Steel.
The significance of this finding is the indication that over the

last decade, Indonesia could have controlled 90% of its industrial
emissions of four key pollutants by focusing on only six sectors of
20
Figure 2.4
IPPS Estimates for Indonesia
Sectoral Share of CO Increase, 1976-1987
(Percent)
30 F
20 -
p
1 131 411
Ii
I
1
10 1213819k
1
212lS9 5
3112 3115 3117312131333219 331134113419 511351336223661 35803611 381338313841 3845
3'3111 31
ISIC Secto
ISIC - International Standard Industrial Classification
Source:
Indonesian Census of Manufactures, 1976 and 1987
Figure 2.5
Sectoral Share of N02 Increase, 1976-1987
(Percent)
30L
25
3813 1
11
1 1
1
1
1
1111
311231163117312131333219331134113419 361136133622 366136603811 38133831 3841 3846
1231
IStCSector

Source:
21
Figure 2.6
Sectoral Share of PT Increase, 1976-1987
(Percent)
311
113
11
3112 31163117 312131333219
341131191611
1 3 1
331134113419 361136133622 356135803811
381338313841 3846
ISICSecr

Source:
Figure 2.7
Sectoral Share of S02 Increase, 1976-1987
(Percent)
I I~
I
I-
I,.,.Ji
19a
_11
DIG
311231163117312131333219 3311341134193611361336223661368038113813383138413846
ISIC
Seto

Source:
production.
22
This has important implications for the current round of
economic and environmental planning.
As policies are drawn up for the
Repelita VI, the next five-year plan, IPPS could be combined with
projected sectoral outputs to identify the sectors that are likely to be
of gravest concern in the years ahead.
If future emissions continue to
be dominated by a few sectors there may be sufficient grounds for

explicitly incorporating this expectation in current planning for
environmental management.
In particular, the institutional structure,
capacity, and mission of the national environmental agency should be

designed to ensure that the priority sectors are favored in the
allocation of resources for the monitoring and enforcement of pollution
control objectives.
In the 1992 World Development Report, World Bank
staff explicitly indicate the importance of favoring these sectors,

stating:
Since all countries face multiple environmental

governments must set priorities on the basis of
so that they can make the most efficient use of
administrative and financial resources. (World
problems,
informed analysis
scarce
Bank, 1992, p. 87)
By aiding in the establishment of priorities, IPPS offers one mechanism

by which governments of rapidly industrializing countries and
development agencies may improve the quality of environmental decisionmaking and enforcement, given their limited resources.
Assessing the Costs of Industrial-Pollution-Control Policies

In the preceding section, I illustrated the use of IPPS to
estimate sectoral emissions of key pollutants and indicated the value of
this information for establishing environmental priorities.
Beyond this
23
relatively straightforward application, IPPS also offers the potential

for a very detailed analysis of the links between environmental policy
and economic outcomes.
Any policy aimed at achieving a given pollution-
control objective will affect the costs of production to some degree;

consequently, it may be expected to affect regional output and
employment.
The objective of industrial-pollution-control policy is to
improve social welfare by accounting for the costs associated with the
externalities of industrial production.
To ensure a net improvement in
welfare, however, the efficiency gains must not be outweighed by any

losses due to reduced levels of production.
This consideration is
particularly significant in developing countries, where the marginal

impact on human health and welfare of a lost unit of production is
likely to be far greater than in the developed world.
This observation
underlines the need to provide policy-makers in developing countries

with a means of assessing the economic impact of proposed environmental
policies.
In this section, I demonstrate the use of IPPS in conjunction with
a regional economic model to assess the economic outcome of a
hypothetical environmental policy.
The objective is to illustrate the
policy issues that may be analyzed given a suitable economic model and
not to make any specific regional policy recommendations.
I have
selected the Commonwealth of Massachusetts to demonstrate this

application because of the ready availability of an appropriate regional
economic model, the Regional Economic Models Inc. (REMI) EDFS-53 model
of the Massachusetts economy.
The REMI model makes dynamic regional
forecasts at a very detailed level, including employment, wages, and
24
output for 53 sectors, personal income, and net regional migration

(REMI, 1992).
As an environmental objective, analysts frequently target total
emissions of a given pollutant, expressed as a percentage of current
emission levels.
For this illustration, I established a hypothetical
objective to stabilize releases of SO2 from manufacturing industries in

Massachusetts at the 1992 level by the year 2005.
objective, I simulated the impact of an
S02
To meet this
tax that met the following
two requirements.
1.
The tax should be levied in proportion to SO2 releases, in order

to provide clear incentives for
2.
S02
The tax should be revenue neutral.
reduction.
All revenue should be
redistributed to the taxed industries, so that no additional tax

burden is imposed on the economy as a whole.
I ran an initial simulation without the tax, in order to establish

a control forecast of the economy and SO2 emissions up to the year 2005.
In subsequent runs, I simulated the tax in the REMI model as a change in
the cost of production.
A percentage change in the cost of production
is equivalent to a percentage tax on output in the REMI model, as total

costs equal the value of output.
In order to create a revenue-neutral
tax, the funds raised were redistributed to the same manufacturing

sectors.
However,
while the tax was raised on the basis of S0 2-weighted
output, it was reimbursed according to unweighted output.
Accordingly,
the net tax rate was higher on the more SO2- intensive sectors.
The
25
requirement that the tax be revenue-neutral significantly complicated

the procedure, as the reimbursement stimulated the economy, increasing
SO2 emissions, while a higher tax depressed output and, as a
consequence, reduced net revenue.
Although I recognized the fact that Massachusetts requires a
balanced budget, due to time constraints, I stopped the model runs at
the sixth iteration, at which point the net tax revenue was negative
$870 million (in 1987 dollars), over the 15-year period.
In other
words, the government was forecasted to spend an average $58 million per
year in excess reimbursements above the amount taken in revenue from the
tax.
Further iterations would have enabled the
S02
target to be reached
with no net expenditure, through fine-tuning of the tax and

reimbursement balance.
Figure 2.8 presents the
S02
tax rates and the
rates net of reimbursement that I used in the last iteration.
The SO2
tax on its own amounts to a levy of $45.68 on every kilogram of S02

released.
The rate of reimbursement was 13.8%, so that the top tax rate
was reduced from 60.0% to a 46.2% increase in production costs, and the
two sectors with no tax received a subsidy that reduced production costs
by 13.8%.
I illustrate the success of this taxation scheme in meeting
the environmental objective in Figure 2.9.
Under the control forecast,
the estimated SO2 emission in 1991 amounted to over 250,000 tons, rising
to 350,000 tons by the year 2005.
With the tax rates shown in Figure
2.8, the SO2 emissions in 2005 were still only 250,000 tons.
Because I
assumed that the sectoral pollution intensities were constant over the
forecast period, I achieved the reduction in SO2 releases solely by
altering the structure of the Massachusetts economy in favor of the less
26
Figure 2.8
Hypothetical 1990 Massachusetts Economy
Simulated S02 Tax and Net Tax with Reimbursement
(Percent Increase in Production Costs)
Two-digit Sic
S02 Tax
Two-digit
SIC
Description
With Reimbursement
S02 Tax
(%)
S02 Tax
Net of Reimbursement
(%)
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Food
Tobacco
Textiles
Apparel
Lumber
Fumiture
Paper
Printing
Chemicals
Petroleum
Rubber
Leather
Non-metallic Minerals
Primary Metals
Fabricated Metal
Non-electric Machines
Electric Equipment
Transport Equipment
Instruments
Miscellaneous Manufacturing
5.90
0.00
18.57
0.00
10.12
2.24
58.02
59.97
34.43
29.06
14.65
8.47
42.02
25.94
10.89
1.17
4.48
0.86
5.91
5.36
SIC - Standard Industrial Classification

Author's calculations based on the World Bank's
Source:
Industrial-Pollution-Projection System, and
the REMI model of Massachusetts
-7.91
-13.81
4.76
-13.81
-3.69
-11.57
44.21
46.16
20.62
15.25
0.84
-5.34
28.21
12.13
-2.92
-12.64
-9.33
-12.95
-7.90
-8.45
27 -
Figure 2.9
Estimated Industrial S02 Emissions in Massachusetts
Control Forecast and Simulation with S02 Tax
(Thousand Tons)
1991
1993
1995
1997
1999
Year
sinulation
+ Control

Source:
Author's calculations based on the World Bank's
Industrial-Pollution-Projection System, and
the REMI model of Massachusetts
2001
2003
2005
28
S02 -intensive sectors.

The true value of combining IPPS with the REMI model in this
analysis is the rich picture that can be drawn of the regional economic
impact of introducing the proposed taxation scheme.
Although the tax
was slightly revenue-negative, the increase in output from those sectors

receiving a net subsidy was not sufficient to off-set the reduction in
output from those suffering a net tax, reducing manufacturing output by
about 2%.
An important aspect of the REMI model is that it allows
people to migrate into and out of the region in response to changes in

personal income and employment opportunities.
Consequently, as
manufacturing output falls relative to the control forecast, reducing

regional employment, regional population falls through net outmigration.
By 1999 the population is predicted to drop more than
employment, so that the crude employment rate (total employment divided

by total population) is actually higher than in the control forecast by
2005.
Although the increased employment, in turn, augments real
disposable per capita income, the income is still about 1% lower than in
the control forecast by the end of the period, despite higher predicted
average nominal wages, because consumer prices are pushed up by 2%
relative to the control.
The rise in consumer prices can be explained by the forecast
change in regional industrial structure.
Eight of the 12 manufacturing
sectors receiving a net subsidy under the SO2 tax and reimbursement
scheme produce durable goods, while six of the eight sectors subject to
a net tax produce nondurables.
Consequently, production of durable
goods increases from 19.7% to 22.7% of total output by 2005, and
29
nondurable production falls from 12.6% to 10.2%.
This narrowing of the
sectoral mix in favor of durable goods not only increases consumer

prices, but also forces the regional economy to become more outwardorientated, with a greater reliance on export markets for durable goods,
and higher imports of the more pollution-intensive nondurable products.
Indeed, this simulation may be seen as a regional example of the muchdebated international phenomenon of the export of "dirty" industries
from developed nations to developing countries.
Analysts already use regional economic models to assess
environmental policies in the United States.
For example a version of
the REMI model has been used by the South Coast Air Quality Management
District (SCAQMD) in the Los Angeles area of California to assess the
socioeconomic impacts of the District's rules and regulations for airquality control.
As I have argued above, this form of analysis is even
more critical in developing countries, where the marginal impact on

individual welfare of job losses or reduced real disposable income per
capita is likely to be higher than in industrialized countries.
In
applying this to Massachusetts, I predicted real disposable income to be

about 1% below the control forecast after 15 years as a result of the
proposed environmental policy.
In the United States, this loss of
welfare might be considered negligible, especially in light of the

anticipated environmental benefits.
In contrast, this loss of income in
a developing country is far more likely to translate into a measurable

loss of health, because the marginal impact on health of a change in
income is likely to be higher.
Depending on the extent of the
detrimental health effect, environmental policy-makers might conclude
30
that it outweighs the health benefits from lower levels of industrial

pollution.
An important criticism of IPPS is that it represents a set of
estimates of pollution intensities prevailing in the United States.
In
the hypothetical application discussed above, I held these intensities

constant,
and I achieved the SO2 target solely by altering the sectoral
composition of the economy.
In other words, I made no allowance for the
effect of technological improvements that reduce pollution intensities.

With an emissions tax of up to 46%, there is little doubt that some
emissions reduction could be achieved at a lower marginal cost than the
rate of tax; the implementation of these improvements would reduce
pollution intensity, and require a reassessment of the appropriate tax
rate for the abating sector.
The assumption of fixed pollution
intensities is especially unrealistic in developing countries, where

considerable improvements in pollution control may be possible.
One
possible adjustment to the IPPS intensities would be to incorporate the

"Penetration Factors" suggested by the World Health Organization (WHO,
1989).
These factors are designed to indicate the effectiveness of
various end-of-pipe pollution-control options.
A feasible ultimate goal
for analysts would be to combine the penetration factors with estimates

of associated marginal pollution abatement costs and, subsequently, link
the costs to a model of the wider economy, in a similar fashion to that
illustrated above.
Such a system would greatly improve the accuracy
with which environmental policy-makers could determine the true costs of

proposed pollution-control strategies.
31
Determining the Benefits of Industrial-Pollution Control

One of the most important goals of environmental policymakers is
to control ambient concentrations of anthropogenically produced
pollutants in order to restrict detrimental impacts on human health and
economic welfare.
The extent of the impact is ultimately determined by
the dose-response characteristics of the target and the pollution dose

received.
The dose, in turn, is a product of the length of exposure and
the mean concentration of the pollutant.
Information regarding the
emission of a given pollutant is most useful to policymakers if it can

be related to an ambient concentration and subsequent impacts.
The World Bank team designed IPPS to estimate emissions of a range
of pollutants from manufacturing activities using widely available
indicators of industrial activity.
Although this design enables us to
conduct some interesting analyses of priority pollutants and sectors and

of the links between economic and environmental policies, as discussed
above, we have no indication of the associated ambient conditions.
IPPS, however, may provide important inputs into a model designed to
estimate ambient concentrations. The following discussion continues the
focus of this chapter on airborne pollutants, although the Biological
Oxygen Demand (BOD), Total Suspended Solids (TSS), and pH data in IPPS
could be used in a similar fashion to model impacts on water quality.
An ambient air-quality model links pollutant emissions to
atmospheric pollutant concentrations.
Such a model involves
considerations of emission patterns, meteorological conditions, chemical

transformations, and removal processes.
The potential complexity of a
comprehensive air-quality model is indicated in Figure 2.10, reproduced
32
Figure 2.10
Components of an Air-Quality Model
Source:
Seinfeld, John H. 1986. Atmospheric Chemistry and Physics of

Air Pollution, Pasadena, CA. John Wiley and Sons, p. 601.
from Seinfeld, 1986.
33
Depending on the required accuracy, and the
desired temporal and spatial resolutions, air-quality models can become

highly data-intensive and computationally complex.
This is particularly
true when the chemical kinetics become a significant factor, which is

the case for primary pollutants with half-lives of less than a few
hours, and for secondary pollutants, such as ozone.
Nevertheless, there
are models available that derive useful approximations for stable

primary pollutants using commonly available meteorological data.
For
example, the World Health Organization's Rapid Assessments procedure

(WHO, 1989) employs a model put forward by the International Institute
for Applied Systems Analysis (IIASA).
Using an inventory of sources
categorized as low-, medium-, or high-level according to the altitude of

the release, and assuming that the dispersion due to wind is equal in
all directions, analysts can compute spatial average and local peak
concentrations.
They can introduce meteorological data as frequency
factors representing the probability of a given combination of

atmospheric stability and wind speed (Dennis, 1978).
The role of IPPS in conjunction with such a model would be to
provide an inventory of industrial emissions from readily available
economic data.
Although this would aid analysts in constructing an
inventory, a number of key sources remain unaccounted for, in

particular:
-
vehicular traffic;
space heating;
municipal incinerators; and
power generation.
Fortunately, there are a number of references that provide standard

emissions factors for such sources, for instance based on the distance
34
driven by vehicular traffic, or the composition of fuel for power

generation (see, for example, U.S. EPA, 1975).
The importance of such
nonmanufacturing sources should not be underestimated, particularly in

modeling urban air quality for developing countries where the
uncontrolled use of biomass and fossil fuels
and poorly controlled
vehicle emissions may be the primary determinants of ambient conditions.

Attempts to model ambient conditions from estimated emissions are
as essential to the formulation of rational environmental policy as
linking emission control policies to the wider economy, as discussed in
the previous section.
An industrial-environmental policymaker's
objective is to improve social welfare by controlling the externalities

of industrial production.
The social benefits associated with the
policies are directly linked to ambient concentrations of key

pollutants.
Importance of Developing Tools for Industrial Pollution Policy-Making

In this chapter, I have demonstrated the essential contribution of
a planning tool, such as IPPS, in the formulation of industrialpollution policy.
At the outset, I framed the discussion in terms of
three components necessary for improving environmental decision-making,

as identified by the World Bank in the World Development Report 1992.
These are namely, better information, stronger institutions, and wider
participation.
I have not only indicated ways in which a tool such as
IPPS may contribute to institutional strengthening and consensusbuilding, but I have also considered in detail the use of IPPS to
improve the quality of information available to planners.
I have demonstrated the use of IPPS to:
Specifically,
35
1.
identify priority pollutants and source sectors;
2.
aid in the analysis of the economic costs of pollution control;

and
3.
help estimate the ambient concentrations of key pollutants, which,

in turn, determine the benefits of controlling industrial
emissions.
In each of these functions, a planner's purpose in using a tool

like IPPS is to improve the rationality, transparency, and certainty of
environmental policy-making.
Rationality is improved by enabling
planners to take into explicit consideration the costs and benefits of a

proposed policy, increasing the chances that the chosen strategy will
lead to an improvement in net social welfare.
This is also likely to
widen the political acceptability of the policy.
Acceptance
additionally depends on the transparency of the process by which

decisions are made, and this too is enhanced through the use of clearly
established models.
The final consideration, certainty, is underlined
by H. Jeffrey Leonard (Leonard, 1985).
The lack of information
available to government regulators generates uncertainty in planning

that creates disincentives for industrial investment.
When directly
monitored pollution data are scarce, systems that allow reliable

estimates to be drawn from more widely available economic data greatly
enhance the certainty of industrial pollution-policy formulation and
implementation, promoting industrial investment and, ultimately,
economic growth.
36
CHAPTER 3
THE INDUSTRIAL-POLLUTION-PROJECTION SYSTEM: INITIAL DEVELOPMENT
Within the World Bank, the need for a method of estimating

releases of industrial pollutants was initially recognized in 1990 by
the Industry and Energy Department of the Research and Policy Vice
Presidency.
The development of the Industrial-Pollution-Projection
System (IPPS) was later continued in the Environmental Assessments and

Programs Department of the same Vice Presidency.
The research team for
the project consisted of three consultants (Mala Hettige, Ralph

Stengren, and the author) and was led by David Wheeler.
Outside the World Bank, interest in methods of estimating
industrial pollution dates back more than 20 years.
Much of the early
work was associated with the development of regional input-output models

to incorporate environmental impacts.
Although most of the work
remained at the theoretical level (for example, Cumberland, 1966; Isard,

1968), some important studies used empirical data relating industrial
production to emissions of pollutants.
Particularly notable in this
regard was Leontief's work in integrating pollution emission and

abatement into a standard input-output technology matrix.
This allowed
him to determine the direct and indirect outputs of industrial goods and
pollution as a function of final demand (Leontief, 1972).
Subsequently,
much interest was focused on the incorporation of pollution estimates

into economic growth models, for example in the Strategic Environmental
Assessment System developed by Resources for the Future in 1975.
More
recently, Faye Duchin et al (Duchin, 1992) developed parameters that use

fuel-composition and combustion-technology data to represent emissions
37
of carbon dioxide, and oxides of sulphur and nitrogen, from combustion

processes.
They then used this information to incorporate emissions
information into a static input-output model of the world economy

designed by Leontief (Leontief, 1977).
In all the studies mentioned above, however, the pollutionemissions data are based on industrial-engineering calculations.
Although standard engineering estimates of pollutant releases based on
mass-balance calculations are available for a large number of processes
and technologies (see, for example, WHO, 1989), the World Bank team
decided to adopt a different approach.
The major problem with
engineering estimates is that they require more detailed information

about the technologies employed than is widely available for most
developing nations.
The World Bank team's objective was to create a
system that would allow us to make reasonable estimates of pollutant

releases from readily accessible national and international databases.
The level of aggregation across industrial sectors in these databases is
generally high.
For example, the United Nations Industrial Development
Organization (UNIDO) annual industrial production data are reported at

the four-digit International Standard Industrial Classification (ISIC)
level, which divides all manufacturing activities into 80 categories.
It is clear that these data have insufficient detail to be used with the
far more detailed engineering estimates.
As an alternative to mass-balance estimates, the World Bank team
chose to use reported emissions from recorded industrial activities.
As
well as being easy to apply to highly aggregated data, an advantage of

this approach is that it uses reports of real emissions, rather than
38
engineering estimates, thus capturing the aspects of production that are

difficult to model, such as the effects of different management
practices.
In the next section of this chapter, I explain the conceptual goal
in using reported industrial emissions to estimate sectoral pollution
intensities.
Subsequently, I give a brief description of the databases
used by the World Bank team, and I discuss some of the more significant
operational complexities encountered in developing the system.
The Conceptual Coal

To the extent that production in a particular manufacturing sector
is associated with a characteristic form and amount of pollution,
analysts can convert information regarding activity in that industry
into an estimate of pollution.
To perform this conversion, they must
multiply the output data by an index of pollution intensity, expressed

as a ratio of emissions per unit of manufacturing activity:
pollution intensity
total emissions
total manufacturing activity
The product of this index and the appropriate measure of

manufacturing activity will then give an estimate of the associated
pollution.
The objective of the World Bank project was to develop a
standard set of pollution intensity indices, for as many manufacturing

sectors as possible.
Analysts can then use (IPPS) to estimate
industrial pollution for any region (country) for which manufacturing

activity data are available.
The World Bank team was able to estimate pollution intensity, as
39
defined above, from the emissions recorded in the U.S. Environmental

Protection Agency (EPA) databases and the measures of industrial
activity recorded in the U.S. Census of Manufactures (CM).
Clearly,
this intensity will vary according to the manufacturing activity.
For
example, the toxic releases from the production of industrial chemicals

are likely to be much greater than those from food production.
Moreover, the production of some chemicals may be more toxic than
others.
This raises the issue of the appropriate level of sectoral
aggregation for the creation of pollution-intensity figures.

As mentioned above, a major objective of IPPS is to allow
pollution estimates to be made from readily available economic data.
Consequently, the World Bank project focused on developing pollutionintensity figures at the four-digit ISIC level of sectoral aggregation,
this being the most detailed and comprehensive level of reporting used
by UNIDO.
To achieve this, we summed the reported emissions by ISIC
sector and then divided by the summed measure of manufacturing activity

within that sector.
This procedure gives an activity-weighted average
pollution intensity, rather than an unweighted average, which would be

obtained if the sectoral pollution intensity was calculated as the mean
of each facility's pollution intensity, as shown below.
Activity-weighted Average Pollution Intensity

P a.
-Z,
(Ei&l
Zi (Ai,)
Unweighted Average Pollution Intensity

PU
E(.'AA
n.
where:
P",
40
= Activity-weighted average pollution Intensity of

Sector s;
P"S
= Unweighted average pollution Intensity of Sector s;
Ei,
= Emission from facility i in sector s;
Ai,
= Measure of activity of facility i in sector s; and
n,
= Number of facilities in sector s.
Over the period 1990-1992, the World Bank team developed

pollution-intensity indices for seven atmospheric pollutants, for
Biological Oxygen Demand (BOD), Total Suspended Solids (TSS), pH and
flow of effluent discharges, and for the toxic releases identified in
the EPA's Toxic Release Inventory (TRI) for the United States.
Although
we could have generated a time-series of pollution intensities, in this

first stage, we only used 1987 data.
We chose this year largely because
it was a census year with consequently detailed CM data.
Building Blocks for IPPS

The World Bank team used a number of EPA emissions databases in
conjunction with data from the CM to develop the IPPS indices.
In the
following sections, I give a brief description of these databases, as

they are important in determining the limitations of the final system.
EPA Emissions Databases

The World Bank team used three major EPA databases in developing
IPPS:
1.
The Toxic Chemical Release Inventory;
2.
The Facility Subsystem of the Aerometric Information Retrieval

System; and
3.
41
The Permit Compliance System of the National Pollutant Discharge

Elimination System.
In addition, we employed the Human Health and Ecotoxicity Database to

provide common units for the chemicals reported in the Toxic Chemical
Release Inventory.
I describe each of these databases more fully below.
The Toxic Chemical Release Inventory (TRI) contains information on

annual releases of toxic chemicals to the environment.
It was mandated
by the "Emergency Planning and Community Right-to-Know Act" (EPCRA) of

1986, also known as Title III of the Superfund Amendments.
The law has
two main purposes: (1) to provide communities with information about

potential chemical hazards, and (2) to improve planning for chemical
accidents.
The TRI requirements cover all U.S. manufacturing facilities that
meet all of the following conditions:
they produce/import/process 50,000 pounds or more of any TRI

chemical, or they use 10,000 pounds or more in any other manner;
they are engaged in general manufacturing activities; and
they employ the equivalent of ten or more full-time employees.
The original TRI requirements, which applied for the 1987 reports, set a
threshold of 75,000 pounds of TRI chemicals produced, imported, or
processed, which was lowered to 50,000 pounds in 1988.
Under the 1987
definition, some 20,000 facilities filed TRI reports, which was reduced
to 18,846 as a result of the delisting of six major chemicals and
increased to 19,762 facilities following the lowering of the reporting
threshold.
The TRI chemicals vary widely in toxicity.
No nontoxic substances
42
or other environmental parameters, such as chemical or biological oxygen

demand (COD/BOD), are recorded.
TRI facilities must report to the EPA
annually all releases of TRI substances to air, water, or land, whether

routine or accidental, and all transfers of TRI substances for off-site
disposal.
Although the identity of a particular substance may be
claimed as a trade secret if justified in advance., only 23 of more than

70,000 TRI reporting forms submitted in 1988 included trade secret
claims.
Quantitative estimates in pounds must be provided for releases
and transfers of TRI chemicals in each of a range of categories,

including:
-
fugitive or nonpoint air emissions;

stack or point air emissions;
discharges to streams or receiving water bodies;
underground injection on-site;
releases to land on-site;
wastewater discharges to Publicly Owned Treatment Works (POTWs);

and
transfers to off-site facilities for treatment, storage, or
disposal.
The World Bank team used the EPA's Human Health and Ecotoxicity Database
(HHED) to provide a common unit of toxicity for the 322 chemicals
reported under the TRI requirements.
The Aerometric Information Retrieval System (AIRS) is the
management system for the U.S. national database for ambient air
quality, emissions, and compliance data.
It is divided into four
subsystems:
1.
the Geo-Common Subsystem, a database of necessary codes;
2.
the Air Quality Subsystem, containing ambient air-quality data;
3.
the Area/Mobile Source Subsystem, which includes estimates of
43
emissions from mobile sources and large-scale, point emission

sources; and
4.
the Facility Subsystem (AFS).
The data used in the creation of IPPS were drawn from the AIRS
Facility Subsystem (AFS).
This contains the emissions and compliance
data mandated by the Clean Air Act that are collected at individual
facilities monitored by the EPA and state agencies.
Compliance data for
over 100,000 point sources are stored in AFS, but emissions estimates
are only available for some of them.
These are generally plants
emitting more than 100 tons per year of one or more of the criteria
pollutants: Particulate Emissions, Carbon Monoxide (CO), Sulfur Dioxide
(SO 2 ),
Nitrogen Dioxide (NO2 ), Lead, and Volatile Organic Compounds
(VOCs).
Although the EPA started collecting air-emissions data in 1973,
they have only entered the information from 1985 onwards into the AFS.
Access to information from years prior to 1985 is more difficult (U.S.
EPA, 1991a).
The Permit Compliance System (PCS) was developed by the EPA to
meet the information requirements of the National Pollutant Discharge
Elimination System (NPDES).
The database contains the self-monitoring
reports of facilities with NPDES permits for discharges of wastewater,

both the permits and the monitoring being mandated by the Clean Water
Act.
Some 60,000 point sources file such reports, based on monitoring
that they perform on a monthly basis; of these there are approximately

50,000 industrial sources, the rest being Publicly Owned Treatment Works
(POTWs).
In the database as a whole, over 2,000 parameters are
44
reported, leading to considerable overlap with the substances reported

for the TRI.
Some of the more important additional parameters are
Biological Oxygen Demand (BOD), Chemical Oxygen Demand (COD), pH, and
temperature.
The length of the time series varies regionally, the
longest being about ten years; however, the data are most complete from
1987 onwards, following the most recent modification of the database
(U.S. EPA, 1990).
Longitudinal Research Database

For industrial activity data, the World Bank team used data drawn
from the Longitudinal Research Database (LRD).
The LRD is an
establishment-level database, constructed from information contained in

the Census of Manufactures (CM) for the years 1963, 1967, 1972, 1977,
1982, and 1987, and the Annual Survey of Manufactures (ASM) for 1973
through 1989.
It is administered by the Center for Economic Studies
(CES), which was set up within the Census Bureau in 1982 to develop the
database, to use the data for the improvement of Census Bureau
operations, and to make the data available to outside users.
The CM is a complete enumeration of all manufacturing
establishments, as classified by the Census Bureau according to the
Standard Industrial Classification System (SIC).
In contrast to the CM,
the ASM is a sample of establishments, selected after each census for

data collection over the following five years.
The annual data
available in the LRD for all establishments from 1972 to 1989 include:
-
establishment name, address, four- and five-digit SIC codes;

payroll statistics, including total salaries and wages;
cost of materials and energy;
capital expenditures; and

total value added.
45
In addition the LRD contains some variables that are only available for
ASM establishments and others that are only collected in census years.
The additional ASM information relates to capital assets, rents,
depreciation, retirements and repair. The data available only for census
years include (1) the quantity and cost of material goods consumed; and
(2) the quantity and value of product shipped.
The product information
collected by the CM (quantity produced, quantity shipped, and value

shipped) is collected at the seven-digit SIC level, which is so detailed
that, on average, each facility reports by three or four product
categories.
Because establishment-level data are collected by the Census
Bureau under the authority of Title 13 of the US Code, the Census Bureau
prohibits the release of information that could be used to identify or
closely approximate the data for an individual establishment or
enterprise.
Consequently, only a limited number of researchers working
as Special Sworn Employees (SSEs) and Census Bureau staff have direct
access to the LRD.
Construction of Pollution-Intensity Indices

Our first task in the development of a pollution-intensity index
from the EPA and LRD data was to match the two sets of information at
the facility level.
This matching was necessary to ensure that the
sector-specific intensities were calculated using emissions and

production data from the same set of facilities.
Unfortunately, there
is no common code to identify the same establishment in the databases,

which necessitated the use of a complex matching process, using the
facility address and Standard Industrial Classification (SIC) code for
46
product sector, with a final alphabetic match on facility name.

lengthy procedure met with mixed success.
This
Of some 20,000 plants
reporting TRI information in 1987, we matched about 13,000 to their

corresponding LRD data for that year.
Emissions data were available on
AIRS for some 20,000 plants, of which we matched about 6,000 to the LRD.
Unfortunately, this figure was cut to 3,000 because not all the
facilities reported emissions for 1987.
The least successful match was
achieved for the PCS; of the 50,000 industrial sources, we only matched
3,000 to the LRD.
Further, when multiple point sources for single
facilities were accounted for, only some 1,500 plants were matched.
Once the matched data-sets had been created, we constructed the
pollution-intensity indices, as outlined above.
Although the concept of
an index of pollution intensity is relatively clear, with pollutant

emission as a numerator divided by manufacturing activity, in practice a
number of questions arose in relation to the choice of numerator and
denominator.
Choice of Numerator
The choice of numerator is relatively straightforward for the
criteria aerometric pollutants, which may all be expressed in units of
mass, as may the Biological Oxygen Demand (BOD) and Total Suspended
Solids (TSS) statistics drawn from the PCS.
Matters are rather more
complicated for pH figures and toxic releases.

Unlike the reported releases of other pollutants, which are
expressed in units of mass, pH is a measure of concentration.
makes a direct summation meaningless.
This
The obvious alternative is to
calculate an average pH for each sector.
This is again complicated by
47
the fact that a neutral pH has the value 7, with acidic and basic pH's
lying on either side of this value.
It is clearly meaningless to take
the average of an acidic and a basic discharge from two facilities

within the same ISIC sector, deriving a more neutral mean discharge, as
this misses the issue of concern.
A final complication is that the PCS
only reports the minimum and maximum pH values of each facility's

discharges, which may themselves fall either side of neutral.
Each of
the pH values also has an associated flow, measured in gallons per day.
The solution adopted by the World Bank team was to calculate flowweighted average minimum and average maximum pH for each sector as shown
below.
Flow-weighted Average Maximum pH

Mx. =
E1
(MaxpH,
ZI MaxFj.
* MaxF2)
Flow-weighted Average Minimum pH

Mn. = Z1 (MinpHi. * MinF,)
Ex MinFi5
where:
Mx.
= Flow-weighted average maximum pH in sector s;
Mn.
= Flow-weighted average minimum pH in sector s;
MaxpHi8
= Maximum pH for facility i in sector s;
MinpH 8
= Minimum pH for facility i in sector s;
MaxFj.
= Flow associated with maximum pH for facility i

in sector s; and
MinFi8
- Flow associated with minimum pH for facility i

in sector s.
We made a final refinement in order to deal with situations in which the
48
set of maximum pHs or the set of minimum pHs for a single sector
contained values on either side of neutral.
In order to make the
results interpretable, we derived the average minimum pH only from the

acidic minimum values and associated flows, unless there were none
within the sector, in which case the minimum basic values were used.
Similarly, we calculated the average maximum pH using only the basic
maximum values and associated flows, unless there were none, in which
case we used the maximum acidic values.
The difficulty with the TRI data arises because the toxicity of
the 322 chemicals reported under the TRI requirements has large
variations.
The list includes, for example, Saccharin and Mercury;
consequently, a simple summation of the total weight of TRI chemicals

released gives little indication of the toxic hazard involved.
Yet, it
would be very unwieldy to develop a separate pollution index for each

TRI chemical.
A better alternative for the comparison of risks is to
weight the releases according to the multi-index categorization of

toxicological potency presented in the EPA's HHED.
We assigned each
chemical's rating under each index to one of three toxicological potency

groups, Group One being the most hazardous.
We also assigned each of
the ten indices to one of five categories of hazard, three of which we

chose for IPPS, as follows:
1.
acute human health and terrestrial ecotoxicity;
2.
acute aquatic ecotoxicity; and
3.
cancer risk.
We distinguished human and terrestrial ecotoxicity from aquatic
49
ecotoxicity because of the significant variation between the

toxicological potency of many chemicals to mammalian and fish life.
For the first two of these indices, we had some difficulty in
converting the ordinal scale ranking of toxicological risk associated
with particular chemicals to a measure of the total risk posed by all
releases from a facility.
The approach we adopted was to multiply the
quantity of each TRI chemical reported by a facility by its

toxicological potency ranking, and then to sum the risk-weighted
quantities for all chemicals released by the facility.
Acknowledging
the questionable validity of using an ordinal scale in an arithmetic

procedure, we used two forms of weighting to test the sensitivity of the
results.
First, we reversed the EPA toxicological potency ratings,
giving a linear weighting scale from 1 to 4.
We used four weights
although there are only three toxicological potency ratings, because

only those TRI chemicals yet to be assigned a toxicological rating were
given the lowest weighting.
Second, we used an exponential weighting
for the four groups, rising by orders of magnitude from 1 to 1,000.
In
a similar fashion, we weighted the releases and transfers of chemicals

posing a carcinogenic risk by multiplying the quantity by the relevant
factor in the Cancer Potency scale.
This factor is an estimate of the
probability of contracting cancer as a result of a lifetime's exposure

to a unit dose.
Because this is a cardinal scale, we did not need to
obtain a separate exponential weighting.
Using this methodology, we
generated five measures of risk-weighted releases and transfers for each

facility:
1.
linear acute human health and terrestrial ecotoxicity;
50
2.
exponential acute human health and terrestrial ecotoxicity;
3.
linear acute aquatic ecotoxicity;
4.
exponential acute aquatic ecotoxicity; and
5.
cancer risk.
In addition, we calculated two unweighted TRI figures for each facility,

these being:
1.
total quantity of TRI chemicals released or transferred; and
2.
total quantity of metal compounds released or transferred.
Although the unweighted total of TRI releases may be useful for a first
level of comparison, analysts should treat it with caution because of
the extreme variation in toxicity between different TRI chemicals.
We
derived the unweighted measure of releases and transfers of heavy metals

to assist in the estimation of the bioaccumulation of metal compounds,
which present significantly different risks to those posed by flows of
other TRI substances.
Choice of Denominator
The LRD provides a number of options for the measure of
manufacturing activity to be used as the denominator in the pollution
intensity indices.
Four of the most obvious are (1) physical volume of
output, (2) shipment value, (3) value added, and (4) employment.
In choosing between them, the need to develop a measure of
pollution intensity that could be used for international comparisons
discouraged us from using either value added or employment, because of
international variations in factor proportions arising.from very
different relative factor prices.
51
Conceptually, the most appealing
choice is physical volume of output, but this poses practical

difficulties.
First, the Census uses a wide range of units to report
output quantities in the LRD even within a given sector, severely

Second, a purely bureaucratic
complicating inter-facility analysis.
problem arises in that many facilities report output volumes in special

samples not included in the main LRD, significantly reducing the sample
size available for analysis.
Finally, the information relating to
output volume in the UNIDO data, the main source for international
comparisons, is not very comprehensive.
Consequently, we used value-of-
shipments as the measure of manufacturing activity in the denominator of

the pollution-intensity index.
Although this statistic has obvious
relative price problems, particularly in the international context, it

has the advantages of a relatively complete UNIDO coverage and the usual
benefit of the dollar metric in allowing inter-sectoral comparison.
Having selected the appropriate numerators and denominators given
the data limitations, the World Bank team calculated pollutionintensities by four-digit ISIC sector.
We used a standard U.S.
Department of Commerce concordance to assign a four-digit ISIC code to

each of the facilities for which EPA data had been matched to LRD
statistics.
The process was complicated by the fact that the
concordance matched five-digit U.S. SIC codes, defining primary product

class, to the four-digit ISIC codes.
This raised difficulties in
dealing with those facilities reporting under more than one five-digit
SIC code, especially when the facility's SIC codes matched more than one
ISIC classification.
The procedure we adopted to deal with this problem
52
was to assign each facility its four-digit ISIC code with the greatest
shipment value.
Although this was generally 80% or more of the total
shipment value, this approach inevitably contributed inaccuracies to the

final estimates of pollution intensity by sector.
Results
Although I critically assess the IPPS indices in the next chapter,
here I briefly describe the outputs produced using the methods detailed
above.
Clearly, the World Bank team could only derive pollution-
intensity indices for those ISIC sectors in which EPA and LRD facility
records were successfully matched.
This sample was further reduced by
the data disclosure requirements imposed by the CES.

these requirements are themselves confidential.
Unfortunately,
Suffice it to say that
a number of four-digit ISIC sectors contained too few facilities to meet

the disclosure requirements; consequently, those matched records could
be used only at a higher level of aggregation (i.e., two- or three-digit
ISIC).
As might be expected, the number of sectors for which we derived
pollution indices varies between the EPA data-bases in relation to the

total number of facility records matched.
This pattern may be
distinguished in Table 3.1 below.

Table 3.1 highlights two issues in particular.
First, as shown in
the third column of the table, we were able to calculate PCS pollution
intensity indices for less than half as many sectors as TRI indices.
Second, the final column in Table 3.1 indicates the number of four-digit
ISIC sectors for which data from more than 50 matched facilities were
available.
As can be seen, of the 34 PCS sectoral indices, only 9 were
calculated using data from 50 or more facilities.
Clearly, with fewer
53
facilities, there is a greater chance that an unrepresentative facility

will bias the calculated pollution intensity.
I cover the implications
of these and other critical issues in the following chapter.
Table 3.1
SUMMARY OF RESULTS
EPA Database
Facilities
Matched to
LRD
4-Digit ISIC
Sectors with
Pollution Indices
Sectors with
more than 50
Facilities
TRI
13,000
74
31
AIRS
3,000
52
18
PCS
1,500
34
Aerometric Information Retrieval System

U.S. Environmental Protection Agency
International Standard Industrial Classification
Permit Compliance System
AIRS
EPA
ISIC
LRD
PCS
TRI
- Toxic Release Inventory
Source: The author's calculations from the Industrial-PollutionProjection System, World Bank.
54
CHAPTER 4
CRITICAL ASSESSMENT AND FURTHER WORK
In this chapter, I examine in some detail the sectoral pollution

intensities for four atmospheric pollutants, from the IndustrialPollution-Projection System (IPPS) derived by the World Bank team using
the methodology presented in the preceding chapter.
After analyzing the
intersectoral variation in the intensities, I discuss some potential

sources of bias in the estimates, arising from the choice of data-bases
and difficulties in linking them together.
Subsequently, I use the
standard deviation of individual plant pollution intensities within each

sector to assess the statistical confidence that may be placed in the
estimates.
The statistical testing raises doubts about the viability of
using pollution-intensity estimates at the four-digit Industrial

Standard Classification System (ISIC) level of aggregation, because for
many sectors there is a very wide range of intensities for plants that
fall within the same sectoral category.
As a first indication of the inaccuracies that are likely to arise
as a result of the uncertainty of the IPPS estimates, I apply IPPS to
the U.S. Census of Manufactures to calculate the total emissions of four
pollutants from U.S. manufacturing.
I then compare the estimates with
the emissions reported by the U.S. Environmental Protection Agency

(EPA).
Although I find one of the four estimates to be within 8% of the
reported emissions, the remaining three pollutant emissions are severely

over-estimated.
Given that some regional variation in emissions
intensities is to be expected, the IPPS estimates are likely to be even
55
less accurate for regional modeling than for national modeling.
To
demonstrate this, I use the system to estimate industrial emissions of

four pollutants in the South Coast Air Quality Management District
(SCAQMD) from economic data in the Census of Manufactures.
I compare
these estimates to the emissions recorded by the SCAQMD, which I find to

be far smaller in magnitude.
Unfortunately, the data available were not
sufficiently detailed to allow a sectoral analysis of the sources of the

over-estimates from IPPS.
It seems most likely, however, that the major
source of error is the wide variation in pollution intensities at the

four-digit ISIC level of aggregation.
I end the chapter by suggesting further work that may be
undertaken by the World Bank team to improve the accuracy of the IPPS
estimates.
This discussion is based on a framework that distinguishes
two distinct levels at which pollution intensity may vary; the

production technology and the end-of-pipe pollution controls.
IPPS Indices
As an indication of the form of results obtained using the
methodology described in the previous chapter, Figures 4.1-4.4 chart the
pollution intensity for four atmospheric pollutants across the 52 fourdigit ISIC sectors for which Aerometric Information Retrieval System
(AIRS) data were available.
The four pollutants are carbon monoxide
(CO), nitrogen dioxide (NO2 ), particulate matter (PT) and sulphur

dioxide
(SO 2 ).
The units of the indices are tons of pollutant released
per million dollars of shipment value.
For ease of presentation, I
assigned the sectors a reference number rather than the four-digit ISIC
code and arranged them in descending order of intensity.
Table 4.1.
- 56 -
Figure 4.1
Pollution Intensity by ISIC Sector
CO - Tons/$ Milion
8070-
1!
6015
50403020
10
10
26 33 1714341 7 2028188163424530412251251395040
37 36 21 19 27 11 15 48 23 4 35 38 13 9 31 43 46 44 29 10 49 6 24 5 47 52
Sector Reference Number
Source:
Figure 4.2
N02 - Tons/$ Milion
90
8070E
.c.
60
-
S50
403020 -
100-
1161Is 4344
12263131472818611313
1126
34 11 17 32 21 28 20 36 4 37 15 35 9 48 44 29 41 42 30 45 46 24 43 50 51 52

Source:
57
Figure 4.3
PT - Tons/$ Million
3025 -
2015 10-
34I131431522Ii5
7 91i212119140123111149I201 2
4131 6 310147
8 28 37 4 17 45 36 32 5 18 16 26 27 48 13 41 29 39 25 50 46 44 38 43 51 52

Source:
Figure 4.4
Pollution Intensity by [IC Sector
S02 - Tons/S Million
60504030-
2010-
211
1513114391251 121618114213140 016 124152
37 17 22 26 27 8 23 18 4 35 19 30 31 10 9 44 2 1 38 16 41 5 47 51 39 49

Source:
Table
4.1
Sector
Reference
Number
58
SECTORS WITH DATA FROM THE AEROMETRIC INFORMATION

RETRIEVAL SYSTEM (AIRS) IN THE INDUSTRIAL-POLLUTIONPROJECTION SYSTEM (IPPS).
Fourdigit
ISIC
Sector Description
Number of
Plants in
Sample
3111
MEAT PRODUCTS
30
3112
DAIRY PRODUCTS
40
3113
PRESERVED FRUITS & VEGETABLES
23
3115
OILS AND FATS
48
3116
GRAIN MILL PRODUCTS
45
3117
BAKERY PRODUCTS
13
3118
SUGAR FACTORIES & REFINERIES
26
3121
FOOD PRODUCTS, N.E.C.
32
3122
PREPARED ANIMAL FOODS
65
10
3133
MALT LIQUORS AND MALT
27
11
3211
SPINNING, WEAVING, & FINISHING
44
12
3219
TEXTILES, N.E.C.
12
13
3231
TANNERIES AND LEATHER FINISHING
17
14
3311
SAWMILLS, PLANING & OTHER WOOD MILLS
15
3319
WOOD & CORK PRODUCTS, N.E.C.
16
3320
FURNITURE & FIXTURES, NONMETAL
104
17
3411
PULP, PAPER, & PAPERBOARD
134
18
3412
PAPER & PAPERBOARD CONTAINERS & BOXES
35
19
3419
PULP, PAPER & PAPERBOARD ARTICLES,
27
20
3420
PRINTING & PUBLISHING
89
21
3511
INDUSTRIAL CHEMICALS EXCEPT FERTILIZER
22
3512
FERTILIZERS & PESTICIDES
32
23
3513
SYNTHETIC RESINS, PLASTICS MATERIALS,
60
24
3521
PAINTS, VARNISHES, & LACQUERS
34
25
3522
DRUGS AND MEDICINES
26
26
3529
CHEMICAL PRODUCTS, N.E.C.
52
146
15
145
(contd.)
Table 4.1
(contd.)
Sector
Reference
Fourdigit
Number
ISIC
59
Sector Description
Number of
Plants in
Sample
27
3530
PETROLEUM REFINERIES
28
3540
MISC. PETROLEUM & COAL PRODUCTS
75
29
3551
TIRE AND TUBES
18
30
3559
RUBBER PRODUCTS, N.E.C.
30
31
3560
PLASTICS PRODUCTS, N.E.C.
51
32
3620
GLASS AND GLASS PRODUCTS
62
33
3691
STRUCTURAL CLAY PRODUCTS
33
34
3692
CEMENT, LIME, AND PLASTER
45
35
3699
NONMETALLIC MINERAL PRODUCTS, N.E.C.
108
36
3710
IRON AND STEEL
191
37
3720
NONFERROUS METALS
85
38
3811
CUTLERY, HAND TOOLS, & GENERAL
13
39
3812
FURNITURE & FIXTURES OF METAL
32
40
3813
STRUCTURAL METAL PRODUCTS
39
41
3819
FABRICATED METAL PRODUCTS
157
42
3821
ENGINES AND TURBINES
18
43
3822
AGRICULTURAL MACHINERY & EQUIPMENT
24
44
3824
SPECIAL INDUSTRIAL MACHINERY &
35
45
3829
MACHINERY & EQUIPMENT, N.E.C.
85
46
3831
ELECTRICAL INDUSTRIAL MACHINERY
39
47
3832
RADIO, TV, & COMMUNICATION EQUIPMENT
30
48
3839
ELECTRICAL APPARATUS AND SUPPLIES,
37
49
3841
SHIPBUILDING AND REPAIRING
20
50
3843
MOTOR VEHICLES
51
3845
AIRCRAFT
24
52
3851
PROFESSIONAL & SCIENTIFIC EQUIPMENT
18
102
103

Source:
The Industrial-Pollution-Projection System (World Bank), and

United Nations Industrial Development Organisation (UNIDO)
ISIC descriptions.
60
links the reference numbers to the ISIC classifications and sector

descriptions and shows the number of plants that were used to calculate
the intensity for each sector.
The most immediately striking feature of Figures 4.1-4.4 is the
approximately exponential distribution of pollution intensities across
the sectors.
I have found this pattern to hold not only for the four
pollutants illustrated, but for all others included in IPPS.
The clear
implication of this distribution is that the total pollution load from

manufacturing may be dominated by releases from a relatively few
sectors.
According to the IPPS estimates, for example, a ton of output
in Sector 33, Structural Clay Products (ISIC 3691) will produce more
than twice as much NO2 as a ton of production in
the next most NO2 -
intensive sector, which is Sector 34, Cement, Lime, and Plaster (ISIC
3692).
Further, the production of cement, lime and plaster is more than
twice as N02 -intensive as the third-ranked sector, Sector 22,

Fertilizers and Pesticides (ISIC 3512).
This provides considerable hope
that problems associated with industrial pollution can be effectively

ameliorated by measures targeted at only a few sectors.
However, it
should be borne in mind that this index does not rank total sectoral
releases, so that it is quite possible for a highly pollution-intensive
sector to have little effect on the total level of releases and
transfers if it has only a tiny share in the local economy.
Despite a few surprises, such as the seventh ranking in NO 2 intensity of Sector 1, Meat Products (ISIC 3111), Figures 4.1-4.4
generally confirm the intuition that the most pollution-intensive
sectors in terms of atmospheric pollution per dollar of output are
61
building-materials, industrial-chemicals, plastics, paper, and metals

industries.
The lowest-ranked sectors are generally those associated
with the production of high-value goods, such as Sector 52, Professional

and Scientific Equipment (ISIC 3851), Sector 51, Aircraft (ISIC 3845)
and Sector 47, Radio, TV and Communications Equipment (ISIC 3832).
Sources of Bias
The methodology used in this study contains a number of sources of
bias that may be reflected in the pollution-intensity results obtained.
I discuss three of the more significant below.
At the very outset, the conditions that must be met before a
facility is required to file a Toxic Release Inventory (TRI), AIRS or
Permit Compliance System (PCS) report (see the previous chapter) impose
two obvious sampling biases, the net outcome of which is unclear.
First, because only those facilities releasing more than a threshold
annual amount of pollution (50,000 pounds of TRI chemicals, 100 tons of
AIRS criteria pollutants, or a
sufficient discharge to warrant a
National Pollutant Discharge Elimination System (NPDES) permit) are

required to report, there will be no record of the very cleanest plants,
even though other plants of a comparable scale of production within the
same sector may be reporting.
This will tend to bias the indices
derived in this study towards over-estimates of average sectoral

pollution intensity.
Second, there may be a number of small facilities
with very high pollution intensities (relative to levels of production)

that nevertheless do not release or transfer more than the threshold
quantity of pollutant.
The omission of these plants will thus
underestimate true mean levels of pollution intensity.
62
The third source of bias arises out of the standard procedure used
to aggregate the 5-digit SIC data to the 4-digit ISIC level.
Under this
procedure, those facilities with Standard Industrial Classification

(SIC) codes that matched onto more than one ISIC code were assigned the
ISIC code with the highest shipment value.
As a result, all releases
and transfers from such facilities were attributed to a single ISIC

code, although in reality some proportion were associated with
activities in a different ISIC category.
It is conceivable that this
approximation might lead to some systematic understatement of pollutant

intensities in the true ISIC sectors, because there may well be scale
economies associated with the higher total shipment value activities
that permit better waste-management practices to be employed.
Statistical Confidence in the IPPS Indices

It is clear from the preceding discussion that there are a number
of reasons why estimates of pollutant releases using IPPS may be
systematically biased, but beyond these methodological considerations
lies the issue of the statistical significance of the IPPS estimates of
sectoral pollution intensity.
In a statistical sense, the IPPS indices
were calculated from a sample of major U.S. emitters, and they are
therefore estimates of the true parameters of that population.
Other
samples from the same population would have produced different

estimates, the scale of the disparity being determined by two major
sources of variation.
First, each four-digit ISIC sector covers a range
of products, processes, and pollution-control technologies, so that the

pollution intensity of individual plants within a sector may be widely
dispersed about the mean.
This would not be a serious concern if each
63
sample contained the same mix of production, as the weighted average

pollution intensity would always be the same.
I expect, however, that
the mix of products, processes, and pollution-control technologies,

measured in terms of share of sectoral output, will vary from sample to
sample; consequently, the estimated weighted average pollution intensity
for each sector will be a random variable distributed about the national
mean.
By analyzing the variation in pollution intensity of the
individual plants in each IPPS sector, analysts can gain some idea of
the probability that the unweighted average IPPS pollution intensities
will approximate the national unweighted average pollution intensity in
that sector.
If the distribution of pollution intensities within each
sector is approximately normal, then I can use the IPPS sample to

specify an interval that will contain the average pollution intensity of
the population with a known degree of confidence.
The interval is
specified as follows:
UP,
where:
(t./
* sd/In)
to
UPS +
(t./
sd/In)
UP,
unweighted average pollution intensity in sector s;
ta/2
the critical value of t for a confidence level of

1-a with n-l degrees of freedom;
sd
= the standard deviation of pollution intensity around

the sectoral unweighted average; and
the number of plants in the IPPS sample in sector s.
The t-distribution can only be used in calculating the confidence

interval if the population from which the sample is drawn is normally
distributed, although the procedure is relatively robust to departures
64
from this assumption (Weiss and Hassett, 1991).
Figures 4.1-4.4
indicate that the distribution of sectoral pollution intensities within

manufacturing is approximately exponential.
I have found this pattern
to be reflected at all levels of sectoral aggregation.
Thus, within the
four-digit ISIC sectors, the pollution intensities of the individual

plants are also approximately exponentially distributed (unfortunately
these data cannot be reproduced here for reasons of confidentiality).
This means that the logarithms of the intensities will be approximately
normally distributed.
Under this assumption the t-distribution is more
correctly applied to the logarithms of the sample values, so that in the

formula above,
UP, is
equal to the average of the log pollution
intensity in sector s, and sd is equal to the standard deviation of the

log intensities.
I used the log intensities of CO, NO2 , PT, and
S02
to calculate
95% confidence intervals, which I transformed into normal form in

Figures 4.5-4.8.
The sector reference numbers are linked to ISIC
classifications in Table 4.1.
There is a 95% probability that the
national unweighted geometric mean pollution intensity for each sector

falls within the range illustrated.
In order to make inter-sectoral
comparisons, I standardized the values, setting the geometric mean of

each sectoral sample equal to a value of one.
The intervals extend
further above unity than below, because they are symmetric in the logs.
For each pollutant in Figures 4.5-4.8, more than half the sectors
have an upper limit to the 95% confidence interval that is more than
double the geometric mean of the sample.
Consequently, there is a 5%
probability that the true value of the geometric mean pollution
65
Figure 4.5
Standardised 95% confidence Intervais

Geometric Mean of cO intensity
g
1
88
98
7-
60
LO
1 3 5 7 9 113 1511719121
6232512712931 331351373941 4314514714951
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52
Note:
Source:
The confidence intervals are calculated from the natural logs

of the pollution intensities of each plant within a sector. These
values are then plotted in exponential form, and standardised
so that the geometric mean of the intensity is equal to unity.
Figure 4.6
Standardised 95% confidence Intervais

Geometric Mean of N02 Irtensity
8
7 -
SS
65 -
43 21 - 0
---
-44
- -
Source:
- -
1 3 5 7
i
11 13 1'511'711'91 2 1
7 291 34513763941 43 45147 49i
51
2 4 6 8 10 121416 18 2022 24 2628 3032 343638 4042"4446 485052
Sector Referenice Number
N2 -Nitrogen
Dioxide
Note:
- -

66
Figure 4.7
Standardised 95% Confidence Intervals
Geometric Mean of PT Intensity
12 11 10 9
8
76
0
L0
78
-E
543-
013 5 911 3 1517192123251 272931331 35137139141 4314514749151

2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48so
52
Note:
Source:

so that the geometric mean of the intensity isequal to unity.
Figure 4.8
Standardised 95% confidence Intervals
Geometric Mean of S02 Intensity
lii
12
1110 9 8 -
7 0
65 4 3 -
2
02
650
135791131171
21
2 7 296 31 335 713941 43145147149i51
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52

Note:
Source:
67
intensity for these sectors is more than double the value calculated
from the IPPS sample.
The most extreme variation is seen in the upper
limit of the PT and SO2 confidence intervals for Sector 12, Textiles
N.E.C. (ISIC 3219), which is more than an order of magnitude above the
geometric mean of the sector sample.
The wide range of the 95%
confidence intervals indicates that at the four-digit ISIC level of

aggregation there is substantial variation in the pollution intensities
of individual plants within sectors.
As a result, it is likely that
there will be considerable inaccuracies in the IPPS estimates of

sectoral emissions.
Testing the IPPS Estimates at the National

and the Regional Level
I have demonstrated above that there is considerable variation in
the pollution intensities of different plants within the same four-digit
ISIC category.
To explore the problems this causes in estimating
pollutant releases for other samples, I used IPPS to model the emission
of four criteria atmospheric pollutants nationally for the United
States, and regionally for the South Coast Air Quality Management
District (SCAQMD).
The pollutants selected were CO, NO2 , PT and
SO 2 .
IPPS Estimates of National Emissions

As I have discussed, the World Bank team calculated the IPPS
intensities from a sample of U.S. manufacturing facilities, the sample
selection being primarily determined by the success of the match between
the EPA and Census of Manufactures data-bases.
I have indicated that
the interval within which the true sectoral pollution intensity can be
expected to fall with 95% probability is relatively wide, for each
sector.
68
In addition, I have discussed some ways in which the sample
selection may be systematically biased.
Both these sources of error are
likely to cause the IPPS estimates of pollution from U.S. manufacturing

to differ from the reported figures.
As an initial test of the extent of these accuracies, I used the
IPPS intensities in conjunction with the U.S. Census of Manufactures to
calculate the total U.S. release of CO, NO 2 , PT and SO2 from
manufacturing in 1987.
I then compared these estimates with the
releases from industrial processes reported by the EPA in the "National

Air Quality and Emissions Trends Report" (U.S. EPA, 1991b).
are illustrated in Figure 4.9.
The results
As can be seen, IPPS is at least
successful in estimating the total emission of PT, over-estimating the

reported emission by only 8%; however, the estimates for the other three
pollutants are all seriously upward-biased.
The estimate for CO is 73%
above the reported release, the SO2 estimate is 149% above the report,
and the NO 2 estimate is 751% above the total reported by the EPA.
Before the true significance of the over-estimates can be
determined, a closer examination of the potential sources of error is
required.
The most optimistic finding would be that the "National Air
Quality and Emissions Trends Report" referred to a smaller set of

facilities, perhaps due to a minimum-size threshold in reporting.
As a
first test of this hypothesis, I compared the emissions totals published

in the report to the totals recorded in the AIRS data used in the
construction of IPPS.
These figures are presented in Table 4.2 below,
together with the IPPS estimates of national emissions.
69
Figure 4.9
U.S. Emissions
IPPS Estimates and EPA Reports (1987)
CO
N02
PT
Polktant
ES IPPS Estimate E EPA Report
EPA - U.S. Environmental Protection Agency


Source:
U.S. Census of Manufactures, U.S. Census Bureau
National Air Quality and Emissions Trends Report, 1991, EPA
S02
70
U.S. EMISSIONS FROM MANUFACTURING
Table 4.2
Pollutant (Million Tons Emitted Per Year)
Data Source
CO
NO 2
PT
SO2
EPA Trends Report1
4.3
0.6
2.4
3.0
AIRS Total2
4.4
2.8
1.3
5.0
IPPS Estimates3
7.5
4.8
2.6
7.5
AIRS
- Aerometric Information Retrieval System
CO
- Carbon Monoxide
EPA
IPPS
- U.S. Environmental Protection Agency

- Industrial-Pollution-Projection System
PT
- Particulates
NO 2
- Nitrogen Dioxide
- Sulphur Dioxide
SO 2
Sources:
1. 1987 emissions from "Industrial Processes" as reported

in: U.S. Environmental Protection Agency, 1991. National
Air Quality and Emissions Trends Report. 1991. Research
Triangle Park, NC. Office of Air Quality Planning and
Standards.
2. Emissions from manufacturing facilities as recorded in
AIRS.
3. Emissions from manufacturing facilities as calculated
from the 1987 Census of Manufactures, using IPPS.
Table 4.2 indicates two important issues.
First, that there are
serious differences between the total emissions from manufacturing

recorded in AIRS and those published in the "National Air Quality and
Emissions Trends Report".
Although AIRS records almost five times as
much NO 2 released as published in the report, the report gives a figure

for PT that is almost double that in AIRS.
The lack of a consistent
direction of discrepancy between AIRS and the report is of particular

concern.
The second issue worth noting is that the IPPS estimates are
consistently greater than the emissions recorded in AIRS.
One possible
71
explanation is that the AIRS minimum emission threshold of 100 tons per
year cuts out a substantial number of manufacturing facilities that are
recorded in the Census of Manufactures.
Alternatively, larger
facilities may have higher pollution intensities than smaller plants,

which would cause the IPPS estimates to be upward-biased, as the
intensities are calculated as size-weighted averages.
At the end of
this chapter, in the discussion of the need for further work, I stress
the importance of examining these issues in more detail.
IPPS Estimates for the

South Coast Air Quality Management District
In the preceding section, I demonstrated that there are
substantial problems with the IPPS estimates of U.S. manufacturing
emissions of four criteria pollutants.
These inaccuracies are likely to
be magnified at the regional level for two reasons.
First, changes in
the composition of output within four-digit ISIC sectors will affect

sectoral pollution intensities.
Second, regional variations in the
regulation of industrial pollution will affect the extent of end-of-pipe

control.
In order to investigate the degree to which these factors
affect the accuracy of the IPPS estimates, I used the system to model
the emission of the same four pollutants from manufacturing industry in
the South Coast Air Quality Management District (SCAQMD).
The SCAQMD is a four-county area around Los Angeles, covering Los
Angeles, Orange, Riverside, and San Bernardino counties.
The local
topography and meteorological conditions have made this air basin

infamous for smog and poor air quality, attracting considerable
legislative attention to the issue of air pollution in the district.
72
The wealth of data generated as a result of this attention makes the

SCAQMD an ideal choice for a case study.
The initial task was to generate the IPPS estimates of atmospheric
emissions by manufacturing industry within the four counties.
In order
to make these calculations, I used the U.S. Department of Commerce (DOC)

1987 Census of Manufactures, Geographic Area Series (U.S. Department of
Commerce, 1990), which gives aggregated shipment values down to the
four-digit SIC level.
I then matched these figures to the four-digit
ISIC classifications used by IPPS.
I encountered a number of
difficulties at this stage, stemming from the DOC disclosure

regulations.
At the geographic level of the county, there are a number
of four-digit SIC sectors that contain an insufficient number of

facilities to pass the disclosure requirements.
Indeed, for the four
counties of the SCAQMD, 35% of the four-digit SIC sectors were prevented
from reporting shipment values.
To overcome this problem, I also collected shipment values by
three-digit SIC.
This reduced the number of nondisclosing sectors to
16% of the total number of sectors in the four counties, but created a
new problem.
At the three-digit SIC level, a number of sectors match to
more than one four-digit ISIC classification, meaning that more than one
pollution intensity may be associated with each sector.
The accurate
solution to this difficulty would be to calculate a weighted average

according to the relative share of each ISIC within the three-digit SIC.
Unfortunately the limited disclosure of more detailed data prevented
this approach.
As a second-best alternative, I estimated two sets of
emissions, the first using the most pollution-intensive ISIC
73
classification for each over-lapping three-digit SIC, and the second

using the least intensive.
This approach thus produced maximum and
minimum estimates of total pollutant releases within each county.
These
are presented in Table 4.3 below, together with the estimates made from
the incomplete four-digit SIC data, and the associated pollution
intensities for each county.
A considerable range may be seen between the high and low
releases estimated from the three-digit SIC data.
In percentage terms,
the smallest increase is 42% from the low to the high estimate of NO2
for Los Angeles county.
The greatest variations are in the estimates
for San Bernardino county, which also has the lowest industrial output.
In this county, the high estimate for NO2 is an order of magnitude
greater than the low estimate.
I am not surprised that the releases
estimated at the four-digit level are generally below the low threedigit estimates, as many more sectors are prevented from disclosing data
at the four-digit level than at the three-digit level.
However, the
estimated pollution intensities at the four-digit level are comparable

to the three-digit estimates, most of them falling between the high and
low intensities.
The objective of creating the estimates of pollutant releases
listed above was to compare them with the reported releases recorded by
the SCAQMD in Appendix III-A to the Air Quality Management Plan, 1991
(South Coast Air Quality Management District, 1991).
For the year 1987,
information on major point sources (those emitting more than 18 tons per
year) is maintained in the SCAQMD's Emissions Inventory System (EIS).
The EIS is updated annually with information submitted by facilities to
Table 4.3
74
ESTIMATED RELEASES AND INTENSITIES FOR THE SOUTH COAST AIR

QUALITY MANAGEMENT DISTRICT.
Estimated Releases
(Thousand Tons per Year)
County
CO
PT
NO 2
SO2
Estimated Intensity
(Tons per $m.)
CO
NO 2
PT
SO2
3-Digit SIC - High Estimates
LA
434.4
149.3
157.2
321.0
4.9
1.7
1.8
3.6
Orange
77.6
26.7
36.4
61.1
3.9
1.3
1.8
3.0
River.
3.6
7.6
7.2
14.2
2.5
5.2
4.9
9.7
San B.
6.3
14.5
18.8
25.7
2.9
6.6
8.6
12.0
3-Digit SIC - Low Estimates
LA
274.2
104.9
36.2
205.7
3.1
1.2
0.4
2.3
Orange
24.8
13.2
5.5
30.2
1.2
0.7
0.3
1.5
River.
1.3
1.2
1.1
4.5
0.9
0.8
0.7
3.1
San B.
0.6
1.2
2.0
2.9
0.3
0.6
0.9
1.3
4-Digit SIC Estimates

165.2
92.6
55.8
196.5
2.5
1.4
0.8
3.0
Orange
11.7
10.4
5.7
27.1
0.8
0.7
0.4
1.9
River.
0.9
3.9
0.8
2.6
1.9
7.8
1.7
5.3
San B.
0.3
0.6
0.8
1.7
0.3
0.7
0.9
2.0
LA
LA
NO 2
Orange
PT
River.
San B.
SO 2
Carbon Monoxide
Los Angeles
Nitrogen Dioxide
Orange County
Particulates
Riverside County
San Bernadino County
Sulphur Dioxide
Source: Author's calculations from the Industrial-Pollution-Projection

System and the 1987 Census of Manufactures.
75
SCAQMD's emission fee reporting program.
By combining these emissions
data with the total value of shipments recorded for each county in the
1987 Census of Manufactures, I was also able to calculate the implied
pollution intensities for each county.
These estimated intensities and
reported emissions are presented in Table 4.4 below.
Table 4.4
REPORTED RELEASES FROM STATIONARY INDUSTRIAL PROCESSES IN

THE SCAQMD, AND IMPLIED POLLUTION INTENSITIES.
Reported Releases
(Thousand Tons per Year)
Implied Intensity
(Tons per $m.)
County
CO
NO 2
PT
S02
CO
NO2
PT
S02
LA
0.3
3.2
15.7
2.5
0.03
0.16
0.03
Orange
1.4
0.8
3.5
0.2
0.05
0.03
0.14
0.01
River.
0.5
0.1
1.7
0.1
0.14
0.04
0.48
0.01
San B.
0.3
0.2
2.1
0.06
0.02
0.35
LA
River.
San B.
*
- Los Angeles
- Riverside
- San Bernardino
- less than 0.005
Source: Reported releases are 1987 data from the SCAQMD, Air Quality
Management Plan, 1991, Appendix III-A (South Coast Air Quality
Management District, 1991). The shipment values used to calculate the
pollution intensities are taken from the 1987 Census of Manufactures (US
Department of Commerce, 1990).
Comparing the estimated releases and intensities in Table 4.3 with

the reported releases and implied intensities in Table 4.4, I show that
the IPPS significantly over-estimates the emissions reported to the
SCAQMD.
This is also reflected in the far higher pollution intensities
calculated from IPPS, even using the low three-digit SIC estimates.
Across the four counties, the low three-digit SIC intensities are more
than two hundred times greater than the implied CO intensities, nearly
76
thirty times greater than the NO 2 intensities, and over four hundred
times greater than the SO2 intensities.
The only pollutant for which
the IPPS intensities are within one order of magnitude of the implied
intensities is particulate emissions (PT), for which the IPPS estimates
are on average double the implied SCAQMD intensities.
The scale of the difference between the IPPS estimates and the
SCAQMD data is an issue of serious concern.
Although IPPS over-
estimated the national emissions discussed in the preceding section, the

scale of the error was less than an order of magnitude.
In contrast,
some of the IPPS estimates for the SCAQMD counties are more than two
orders of magnitude above the reported emissions.
may be given for the size of the error.
A number of reasons
First, it may be that the mix
of products and processes within each four-digit ISIC sector is

substantially different from the mix in the sample used to calculate the
IPPS indices.
In the preceding discussion of the statistical
significance of the IPPS intensities, I demonstrated that there is

considerable variation in the intensities of individual plants within
the same sector.
Consequently, if stricter pollution controls in the
SCAQMD formed a significant disincentive for the location of more

pollution-intensive plants in the Basin, thus affecting the mix of
products and processes within ISIC sectors, the sectoral pollution
intensities might be severely affected.
The second possible explanation for the IPPS over-estimate of
SCAQMD pollution is related to the first.
Even if the SCAQMD mix of
products within each ISIC sector is not significantly different from the
IPPS sample, it is possible that the stricter control of air pollution
77
in the SCAQMD has pushed the pollution intensity of each plant well
below the national average for the particular product.
Although the Los
Angeles basin is one of the most polluted air-sheds in the United

States, this is largely due to meteorological conditions and motorvehicle emissions (SCAQMD, 1991).
The need to reduce the high ambient
concentrations of criteria pollutants has led to regulatory controls

over industrial emissions that are in many cases more stringent than the
national average (SCAQMD, oral communication).
A more subtle explanation for the error in the IPPS estimates of
the SCAQMD emissions is related to the way in which the IPPS pollutionintensity coefficients were calculated.
As I explained in Chapter 3,
the IPPS coefficients are size-weighted average pollution intensities.

Consequently, the pollution intensities of the largest plants in each
ISIC sector of the World Bank team's sample have a greater influence on
the IPPS coefficents than the pollution intensities of the smallest
plants.
If there is a systematic relationship between plant size and
pollution intensity, then the size-weighted coefficients will

misrepresent the average pollution intensity of samples with a different
mix of plant sizes.
Unfortunately, the data available did not allow me
to compare the SCAQMD mix of plant sizes with those in the World Bank
team's sample.
For the sake of completeness, I note that it is also possible that
the error in the IPPS estimates is partially attributable to errors in
the recording of SCAQMD emissions.
In the previous section I identified
significant differences between the national emissions reported in AIRS

and those presented in the "National Air Quality and Emissions Trends
Report".
78
This indicates that errors in the officially recorded
emissions cannot be discounted.
Further Work
The analysis that I have presented in this chapter has raised some
important issues regarding the use of the IPPS indices.
In particular,
it is essential that further effort be devoted to uncovering the true

sources of the disparity between the IPPS estimates and the national and
SCAQMD reports.
A likely conclusion is that there is too much variation
in the factors affecting pollution intensity at the four-digit ISIC

level of aggregation to place much confidence in a single estimate of
the sector's pollution intensity.
In order to improve the accuracy of
the IPPS estimates, further work is needed to identify and control for
the sources of variation.
Although substantial effort would be required
to examine all sources of variation in all sectors, there is hope that

significant improvements in the IPPS estimates can obtained by focusing
future work on a few important sources of variation in the most
pollution-intensive sectors of production.
The exponential distribution
of pollution intensities illustrated in Figures 4.1-4.4 indicates that

the pollution load from a set of manufacturing plants is highly
dominated by production in the most pollution-intensive sectors.
Analysis of the key sources of variation that I suggest below should
focus first on these critical sectors.
One of the major problems with the current set of IPPS
coefficients is that each four-digit ISIC sector covers a number of
production processes, each with a different level of pollution
intensity.
The existing sample of matched EPA/DOC data could be used to
79
test the improvement in IPPS estimates at more detailed levels of

aggregation, ultimately down to the seven-digit level of SIC product
classification.
At this level of product definition, there is only
limited variation in production technologies.

two problems.
This approach will raise
First, it will not be immediately possible to generate
more detailed disclosable pollution-intensity estimates for a large

number of more finely defined sectors.
Using the existing IPPS sample
of matched EPA/DOC data, any increase in the number of sectors reduces

the number of plants in each sector, meaning that fewer sectoral indices
pass the Bureau of Census disclosure requirements.
Nevertheless,
improvements in the IPPS estimates with nondisclosable data can be

tested on the premises of the Bureau of the Census, and disclosable
estimates can be developed for those sectors with a larger number of
plants.
Second, the four-digit ISIC categorization was justified as an
appropriate level of aggregation on the basis that this is the most

common degree of aggregation for economic data from developing nations.
The immediate ease of application of IPPS would be substantially
curtailed if it became necessary to develop and apply IPPS indices at a
more finely detailed level of industrial classification.
There is some
hope, however, that the dominant effect of a few highly pollutionintensive sectors may mean that substantially improved estimates can be
obtained with more detailed classifications from a relatively few key
sectors.
At a sufficiently detailed level of product definition there is
only limited variation in the available production technologies.
Variation in pollution intensities is therefore primarily determined by
80
end-of-pipe pollution-control measures.
The inclusion of this source of
variation would greatly improve the accuracy of IPPS estimates, but

raises two difficulties. First, how to relate information regarding
pollution-control measures to emission intensities, and second, how to
determine what degree of pollution-control is employed by the facilities
being modeled.
The first issue is more easily solved than the second,
as extensive data are available on the effectiveness of pollutioncontrol technologies.
For example, the World Health Organization's
Rapid Assessment procedure (World Health Organization, 1989) provides

"Penetration Factors" for a number of different stages of waste control
for both emissions to air and effluent discharges, expressed as a
percentage reduction of the uncontrolled release.
Alternatively, the
Pollution Abatement and Control Expenditures (PACE) data collected by

the DOC could be used in a regression analysis to relate pollutioncontrol costs to pollution intensity.
Determination of the degree of pollution control employed by a set
of manufacturing plants to be modeled using IPPS is more difficult.
Such pollution-control data are not widely available in developing
countries, although it is generally assumed that the degree of control
is less than in most industrialized nations.
One approach to dealing
with this problem is suggested by the Industrial Efficiency and

Pollution Abatement Project (IEPA), commissioned jointly by the World
Bank and the Indonesian Ministry of Industry in 1991 (World Bank, 1991).
This data collection study surveyed a number of key polluting sectors to
determine the degree of pollution control, broadly categorized as
"Western Standard", "Below Western Standard" and "None".
Clearly, a
81 -
trade-off has to be made between the accuracy of the adjusted IPPS

estimates and the resources expended in data collection.
Although process technology and end-of-pipe pollution controls
are likely to account for much of the variation in pollution intensity
within the four-digit ISIC sectors, additional factors may play some
role.
age.
In particular, there may be some effect due to plant size and

This hypothesis can be tested relatively easily with a regression
analysis of the production and investment data recorded in the LRD.
The
most difficult aspect of the test will be to control for the effect of
differences in process technology and end-of-pipe controls.
The suggestions I have made above represent a substantial program
of further work aimed at improving the accuracy of the IPPS estimates,
but I believe that the potential of such a system to improve the quality
of environmental decision-making for developing countries justifies the
additional investment of time and resources.
82
BIBLIOGRAPHY
Bhagwati, J. 1992.
Letter to The Financial Times.
14 February, 1992.
Cumberland, J.H. 1966. "A Regional Inter-industry Model for the

Analysis of Development Objectives." Regional Science Association
Papers, Vol. 16, p. 69.
Dennis, R.L. 1978. The Smeared Concentration Approximation Method:
Simplified Air Pollution Dispersion Methodology for Regional
Analysis. Laxenburg, Austria: International Institute for
Applied Systems Analysis.
Duchin, F., Lange G., Thonstad K., and Idenburg A. 1992. Strategies for
Environmentally Sound Economic Development. New York, NY. Report
for the United Nations.
Isard, W.E. 1968. "On the Linkage of Socio-Economic and Ecologic
Systems."
Regional Science Association Papers, Vol. 21, p. 79.
Leonard, H. Jeffrey, 1985. "Confronting Industrial Pollution in Rapidly
Industrializing Countries: Myths, Pitfalls, and Opportunities."
Ecology Law Quarterly, Vol. 12, p. 779.
Leontief, Wassily W. and Ford, D. 1972. "Air Pollution and the Economic
Structure:
Empirical Results of Input-Output Computations."
In
A.P. Carter and A. Brody, eds., Input-Output Techniques. NorthHolland Publishing Company.
Leontief, Wassily W., Carter A. and Petri P. 1977. Future of the World
Economy. New York, NY. Oxford University Press.
REMI. 1992. Model Documentation for the REMI EDFS-53 Forecasting and
Simulation Model. Vol. 1. Amherst, MA: Regional Economic
Models, Inc.
Seinfeld, John H. 1986. Atmospheric Chemistry and Physics of Air
Pollution. Pasadena, CA. John Wiley & Sons.
South Coast Air Quality Management District. 1991. Air Quality
Management Plan. Los Angeles, CA. South Coast Air Quality
Management District.
Summers, L. 1992.
Quoted in The Economist. 8 February, 1992, p. 66.
Tobey, James A. 1989. "Economic Development and Environmental

Management in the Third World." Habitat International, Vol. 13,
No. 4, pp. 125-135.
U.S. Department of Commerce. 1990. 1987 Census of Manufactures,
Geographic Area Series. California. Washington, DC: Bureau of the
Census.
83
U.S. Environmental Protection Agency, 1975. Compilation of Emission

Factors, U.S. Environmental Protection Agency. AP-42, Part A.
Washington, DC: U.S. EPA.
U.S. Environmental Protection Agency. 1990. Permit Compliance System.
Office of Water Enforcement
Executive Summary. Washington, DC:
and Permits.
U.S. Environmental Protection Agency. 1991a.
Volume I. Research Triangle Park, NC:
AIRS User's Guide.

National Air Data Branch.
U.S. Environmental Protection Agency. 1991b. National Air Quality and

Emissions Trends Report, 1991. Research Triangle Park, NC:
Office of Air Quality Planning and Standards.
Weiss, Neil A. and Hassett, Michael J. 1991.
Reading, MA: Addison Wesley.
Introductory Statistics.
World Bank. 1991. Industrial Efficiency and Pollution Abatement (IEPA)

Project Data Collection Study. Internal Report.
World Bank. 1992. World Development Report 1992: Development and the
Environment. New York, NY: Oxford University Press, Inc.
World Health Organisation. 1989. Management and Control of the
Environment. Limited Distribution.
84
GLOSSARY OF ABBREVIATIONS
AFS
AIRS Facility Subsystem
AIRS
Aerometric Information Retrieval System
ASM
Annual Survey of Manufactures
BOD
Biological Oxygen Demand
CES
Center for Economic Research
CO
Carbon Monoxide
COD
Chemical Oxygen Demand
DOC
Department of Commerce
CM
Census of Manufactures
EPA
U.S. Environmental Protection Agency
HHED
Human Health and Ecotoxicity Database
IPPS
The Industrial Pollution Projection System
ISIC
International Standard Industrial Classification
LRD
NO 2
Nitrogen Dioxide
NPDES -
National Pollutant Discharge Elimination System
PCS
Permit Compliance System
PT
Particulates
REMI
Regional Economic Models Incorporated
SCAQMD -
South Coast Air Quality Management District
SIC
Standard Industrial Classification
SO 2
Sulfur Dioxide
TRI
Toxic Release Inventory
TSS
Total Suspended Solids
UNIDO -
United Nations Industrial Development Organization
VOC
Volatile Organic Compound
WHO
World Health Organization

The Industrial Pollutionm

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

The Industrial Pollutionm

Încărcat de

Drepturi de autor:

Formate disponibile

THE INDUSTRIAL-POLLUTION-PROJECTION SYSTEM:

CRITICAL ANALYSIS AND POTENTIAL APPLICATIONS

SUBMITTED TO THE DEPARTMENT OF URBAN

@ Paul J. Martin 1993

- 2 THE INDUSTRIAL-POLLUTION-PROJECTION SYSTEM:

Thesis Supervisor: Professor Karen R. Polenske

I wish to express my gratitude and appreciation to all those who

Many people gave freely of their time and

energy to provide the necessary data, for which I would particularly

Credit for the creation of the

Valuable criticism and counsel have also come from

Professor Lawrence Susskind, for which I am very grateful.

Obstacles to Effective Industrial-Pollution-Control Policy...... 7

Assessing the Costs of Industrial-Pollution-Control Policies... 22

In Bangkok, Beijing, Calcutta, Jakarta, Tehran, and other urban

The resulting sickness and death is worse

than would be expected under the same conditions in developed nations,

Between 300,000 and

Other priority pollution issues in the

world's rapidly industrializing regions include the exposure of more

In this thesis, I examine

the development by the World Bank of the Industrial-Pollution-Projection

importance of such a tool for improving the quality of environmental

In order to place these

issues in some context, I outline below some of the obstacles facing

- 7 effective industrial-pollution policy in developing countries.

failure arises because incomplete information and transaction costs

As a result, net social

The social costs of pollution take a number

of forms, which may be summarized in three broad categories as (1)

Although detriments to human health may also

cause a reduction in productive capacity, they are defined separately

The definition of lost productive capacity

includes damage to both physical capital and natural resources.

aesthetic utility includes the many aspects of individual enjoyment of

In practice, the assessment of both the costs and benefits is

The controversy surrounding this task is highlighted by an

Despite the well-recognized detrimental impacts of

industrial pollution outlined above, Summers wrote:

Understandably, this memorandum caused an international uproar.

Perhaps the most obvious social

costs of pollution are those associated with adverse impacts on human

Just as the marginal detriment to health is likely to be higher

in developing countries because of pre-existing ill-health and

Bhagwati wrote in response:

- 9 If part of the aim of industrial-pollution control is to improve

By converting the abatement

of analysis is complex, and it can be conducted only in approximate

Without reliable information, policy-makers

Without sufficient institutional capacity, the information

cannot be effectively analyzed and the policies cannot be implemented.

The development of IPPS by

the World Bank was undertaken in recognition of these needs.

IPPS is intended to allow better-informed policy-making by providing

Second, IPPS is intended to

strengthen institutional capacity in two ways.

The system will allow

more efficient use to be made of scarce financial and administrative

pollution-control policy, IPPS will improve the transparency of

This is especially important for government and

development agencies facing political pressure in this arena.

I then demonstrate how IPPS may be

used in conjunction with economic models to assess the costs of an

I critically assess the